Optical resolution - measured in dots per inch (dpi). A characteristic showing the higher the resolution, the more information about the original can be entered into the computer and subjected to further processing. A characteristic such as “interpolated resolution” (interpolation resolution) is often given. The value of this indicator is doubtful - it is the conditional resolution to which the scanner program “undertakes to count” the missing points. This parameter has nothing to do with the scanner mechanism and, if interpolation is still needed, then it is better to do it after scanning with a good graphics package.

Color depth

Color depth is a characteristic that indicates the number of colors that the scanner can recognize. Most computer applications, excluding professional graphics packages such as Photoshop, work with 24-bit color (16.77 million total colors per dot). For scanners, this characteristic is, as a rule, higher - 30 bits, and for the highest quality flatbed scanners - 36 bits or more. Of course, the question may arise - why should the scanner recognize more bits than it can transmit to the computer. However, not all received bits are created equal. In CCD scanners, the top two bits of theoretical color depth are usually “noise” and do not provide accurate color information. The most obvious consequence of "noise" bits is insufficiently continuous, smooth transitions between adjacent brightness gradations in digitized images. Accordingly, in a 36-bit scanner, the “noise” bits can be shifted far enough, and in the final digitized image there will be more pure tones per color channel.

Dynamic range (density range)

Optical density is a characteristic of the original, equal to the decimal logarithm of the ratio of light incident on the original to reflected light (or transmitted light for transparent originals). The smallest possible value 0.0 D is a perfectly white (transparent) original. A value of 4.0 D is a completely black (opaque) original. The dynamic range of the scanner characterizes what range of optical densities of the original the scanner can recognize without losing shades either in the highlights or in the shadows of the original. The maximum optical density of the scanner is the optical density of the original, which the scanner still distinguishes from complete darkness. All shades of the original darker than this border cannot be distinguished by the scanner. This value very well separates simple office scanners, which can lose detail, both in dark and light areas of the slide, and, moreover, in the negative, from more professional models. Typically, for most flatbed scanners, this value ranges from 1.7D (office models) to 3.4D (semi-professional models). Most paper originals, whether photographs or magazine clippings, have an optical density of less than 2.5D. Slides require, as a rule, a dynamic range of more than 2.7 D for high-quality scanning (Usually 3.0 - 3.8). And only negatives and X-rays have higher densities (3.3D - 4.0D), and it makes sense to buy a scanner with a high dynamic range if you work mainly with them, otherwise you will simply overpay the money.

In general, the idea of ​​scanning and organizing old photographs, of course, was hatched for a long time, it is not easy to decide on such a volume of work on scanning old photographic films (more than a hundred) and photographs (thousands). In general, since childhood, he wanted me to have digitized old photographs of great-great-great-grandmothers, and finally, 20 years later, he decided to move on to this business.

Scanner

The first question was - of course the scanner. At one time, about 7 years ago, I tried to digitize negatives and decided to stock up on a film scanner. There was not much money, I chose which is cheaper, it turned out to be Miktotek Filmscan 35.


Compared to the monsters of scanning, it cost a penny, but the result was terrifying. I used Silverfast for it as the most advanced software at that time (maybe now). I don't know why, but sometimes with different passes this miracle gave me a blue or a green photo, then everything hung, it was unpredictable and very sad, I had to pore over each frame for 10-15 minutes, straightening the histograms and performing other dances with a tambourine. In general, this process discouraged me from scanning films for several years, the scanner is lying around somewhere.

Now, having considered all the pros and cons, the following was decided.
There were several points to consider:

• for the most part, it will not be me who will scan, but the parents, since they have time now
• you need to scan not only films, but also photos
• you need to scan a lot
• no fabulous budget

In addition to all the above, I understood that now the film is no longer an actual medium, and therefore most likely it will be necessary to scan only once, although it may take a lot of time.

So, film scanners fell away for two reasons:
Firstly, previous experience has shown that you cannot buy such a unit for a cheap normal unit, but what is cheap - oh, I can't stand such hell a second time.
Secondly, buying a separate scanner for photos and separately for film is also somehow expensive and impractical.
Moreover, I said to myself, if I come across something good, I will take it to a professional laboratory, for a dozen shots you can go broke.

Having looked at what is on sale that can scan film, in addition to paper, it turned out that the choice is not great: either, again, sky-high prices, or just a couple of options. After the break of all shops that were open immediately after the holiday, it turned out that there are the following acceptable options:

• Epson Perfection V330 Photo (A4, 4800 x 9600 dpi, USB 2.0, CCD, Film Adapter)
• Epson Perfection V370, Photo (A4, 4800x9600 dpi, CCD, USB 2.0)
• Canon CanoScan LiDE 700F (A4 9600х9600dpi 48bit CIS Slide Adapter USB2.0)
• Canon CanoScan 5600F (A4 4800x9600dpi 48bit Slide Adapter USB2.0)

The rest was either too expensive, from 10,000, or, conversely, nothing was skillful. Unfortunately, the CanoScan 5600F was dropped due to the lack of it at the moment on sale, although according to the description it is very good. The rest turned out, according to reviews, about the same, but the decisive role was played by the fact that Epson "s had drivers for Linux, and since I would like to work not only under Windows, in the end the Epson Perfection V330 Photo won. What makes the 330 model different from the 370, but since Linux drivers were mentioned only for 330, I settled on it, so to speak, "in order to avoid".

Unfortunately, I didn't have time to try it for Linux yet, but in the Windows software I liked the function of removing defects - it works with a bang on black and white old photographs. But you also need to be careful with her - sometimes something worthwhile can be considered a defect.

In the reviews about the scanner, in some places the problem with the appearance of stripes when scanning films is mentioned - but I have not seen this yet. Nevertheless, in my opinion, here is something useful about this, found in one of the reviews on the Yandex market: “Two years later, I can report on the results of the investigation: there is a calibration window in the scanner frame where the white balance is set. If specks of dust get there, "broken pixels" are obtained, which, when the carriage is run, give stripes. This is most likely a design defect in the new LED backlight(but who will admit it ...). So gentlemen, if you have such a scanner,
remove dust. "

With what resolution to scan - this question was not the last. The scanner produces a maximum of 4800x9600, but when trying to set this when scanning a 9x13cm photo, the system began to swear at the scale, I had to reduce it.

The criterion for choosing a resolution is simple: if you consider that you can print with a standard resolution of 300dpi, then in order to get the same image, you need to have a minimum of 300dpi. Considering that the photos are old, then there is no point in overestimating this figure - all the same physical resolution will not allow you to get quality out of nothing. Again, hardly anyone would ever want to print a poster with a picture of their great-grandfather in A1 or even A4 format. If anyone writes a book, it is unlikely that the picture will be larger than a sheet. In general, I decided that for very old ones a two-fold excess will come down, for better and later ones - a three-fold excess, i.e. 600dpi and 900dpi respectively. Then I chose what was the closest to what the software gave out that came with the scanner.

I decided to use the maximum for negatives - it was not in vain that I bought with such a resolution ... Most likely this is an overkill of 4800x4800dpi, but you can always cut it down later, but the main thing is that you don't have to rescan with other parameters and you can sleep peacefully.

Scans are saved, of course, in no case in jpeg, in order to avoid compression losses. Everything is just tiff. It seems, of course, that the place eats more, but then scan it once - and then you don't know the problems: I do what I want. I also did not come to this right away, but practice shows that if I save now, then I will regret and return to this issue, and so, if everything is to the maximum, then there is nothing to regret later.

Cataloging

Naturally, after digitization, the whole thing needs to be raked somehow. The main task was to sign great-great-relatives, because I wanted to save the history of the family for the future, and no one will ever figure it out without competent comments.

The option to immediately process photos and upload them to the site was not suitable for two reasons: firstly, you need to process everything at once, and this is time, and the parents do not understand anything about this; secondly, technologies are changing, and who would have known how the site would look like in a couple of decades, if it would exist at all.

The use of a smart cataloging program was not suitable for the same essential reason - there is no guarantee that in a few decades this software will be alive and, accordingly, no one will understand what, where and how is stored in its smart unique format.

The decision came to mind to keep the description in the usual text file with the same name as the photo - the text, and in Africa, the text will probably be read by anyone after ten decades, even if some other super-unicode is invented, nevertheless it is much more reliable than special software. But as a programmer, I looked at this option with horror - well, it was ugly and that's it. Yes, and inconvenient in the process.

Parents said that they generally want it like in a Word - here's a photo, here's a signature - and everything is clear. From such a proposal, the hair stood on end, for again - today there is a Word - tomorrow it is not.

Another option is to store signatures in EXIF. It was embarrassing that when processing pictures, many EXIF ​​softwares are simply ignored, as a result, losing precious signatures may be irreplaceable.

In general, after analyzing the whole situation, I made a decision: we scan the photo, sign it in the form of EXIF ​​and then do all these pictures with signatures read-only, so that there is no temptation to change something, and thus we guarantee the safety of information. If you want to change - make a copy - and go. Well, backups, of course. And in general, in the end, that's what we programmers are for, in order to sketch out a small script so that the entire EXIF ​​could be exported to a text file, just in case, "in order to avoid" :)

There are a bunch of tools for working with EXIF ​​in Linux command line but this is not acceptable for comfortable work with a lot of pictures. However, here's what it is: exif, exiftool, exiv2, google it for more information. Next, I used exiftool for batch processing, but more on that later.

Let's see what is from the GUI. After studying what the OpenSource community offers us, I somehow settled on DigiKam - “digiKam is an advanced digital photo management application for Linux, Windows, and Mac-OSX,” as they say on their website.
I decided to edit in GIMP, GNU Image Manipulation Program, similar to Photoshop, but opensource. Therefore, the ability to edit photos for the cataloging software was not required separately, but several things were bribed in cataloging itself.

First, DigiKam edits EXIF, which is what I need.

Secondly, all the photos are immediately on the screen, we sign in the window next to it and immediately go to the next one - quickly, simply and conveniently.

Thirdly, it was noticed that in EXIF ​​itself there are several similar tags for commenting: Comment, UserComment, ImageComment So, DigiKam writes to everything at once, so the probability that this information will be read by other software is quite high.

In addition, reading the reviews, I was pleased with the thought that apart from just EXIF, the software can maintain a catalog, and without copying anything anywhere, unlike many others, but simply processing everything on the spot. This was a huge plus - I did not look for this opportunity initially, but it turned out to be very useful. And what I liked - besides entering information in EXIF, she writes it to her database and then it is convenient to sort the photo and search by tags, tags, descriptions, etc. And even if at some point the software disappears and the database also disappears, then a copy of the data will remain in EXIF, which, in fact, is what I need.

Some interesting ideas on cataloging are described in the already mentioned article “Experience of creating a catalog and indexing a family photo archive. Indexation and Digitization of Films ". So, all or almost all of this data can also be kept in EXIF ​​and, if necessary, exported to any format, as it is convenient for us.
An additional advantage of DigiKam is that you can choose any photo as the cover of the album, and I liked the idea of ​​having a photo of the paper album itself as the cover, for which thanks to the author.

Another unobvious moment that I encountered when working with DigiKam: if there is no permission to write to a photo file, then the software silently writes only to its database, not making it clear that there are problems. For a long time I tried to figure out why there is a signature in the program, but not in the file, especially since the "save to file" option is set in the settings. So, keep this in mind - check the access rights, otherwise you can swear for a long time.

We put it on the site

So, the main tasks have been solved - scanning and cataloging. Now it's time to boast to your relatives, show your friends the photo. Naturally, by uploading photos to the site. Not so long ago I already made a softphone for this case: I put the necessary photos in
catalog, launched - and ready, the album was made. I wrote about this on Habré last time, "Simple automation: photo album". Now, using DigiKam, I decided that you can mark a photo directly in EXIF ​​tags, whether it should be placed in a photo album or not, because during scanning there were all sorts of pictures that should not be posted on the site. And comments can now be taken from EXIF.

Everything seems to be good, but not very good.

Everything on the site is processed in PHP, and there is, as it seemed to me, a wonderful function for working with EXIF, read_exif_data (), but as practice has shown, this under-function shows only part of the data, completely silent about the rest. I rummaged through everything I could - and the dream of an easy life sank into oblivion, I had to pull EXIF ​​out of the files at the stage of album generation, since there are command line tools.

As a result, I rewrote the script, recalling the caustic commentary on my previous article "The generator of php files in Perl ... Monsieur knows a lot ..." now the leg, and so a couple of minutes - and the problem is solved.

So, when processing a photo in DigiKam, we mark the photo with a flag (it is called PickLabel there). The flag is written to the file in EXIF. When we process all the files from the directory, we pull out the checkbox using exiftool:

$ flagPickLabel = `exiftool -b -PickLabel" $ fname_in "`;

Well, then, depending on the checkbox - if it is, then we process it, if not, we skip it. Everything is set on the command line, so that it is convenient. In fact, here you can process a lot of things, it already tastes and colors whoever needs it.

Link to the source code, in case someone suddenly needs to look closely or even apply: photo_album-r143.tar.gz. How to use - mentioned in the previous article, I will not repeat myself.

Thank you for your attention, and if someone came in handy, then I'm immensely glad.
Criticism is welcome.

UPD: I accidentally found it on Habré about scanning negatives - I'm surprised I didn't notice it before. Let it be here to the heap.

Tags:

Add tags

Since in this article we will be talking exclusively about scanning transparent originals - slides and negatives - I will omit all discussions about opaque samples. The article is written for a reader trained in photography and computer image processing, as well as knowing the basic concepts: optical density interval, useful optical density interval, latitude of photographic material, contrast, average gradient, etc.

What do we have?

D First, let's take a look at the settings for the Epson Perfection 1650 photo scanner. He is the only one I have, and it would be strange if I were describing something else. So, according to some data, this scanner in the mode of scanning a transparent original can perceive the density difference ΔD scanner = 3.2, according to other data, its dynamic range is ΔD scanner = 3.0. My research shows much more modest characteristics in this parameter, therefore, manufacturers are cunning (although they do not indicate the dynamic range at all, at least for scanners of this level), saying that we can "painlessly" scan a color negative. I argue that in the form in which the scanner is supplied, a lossless color negative cannot be scanned. So let's get started.

What do all these letters, numbers mean?

D- density, or decimal logarithm of opacity. It is known that the human eye perceives a scale that uniformly increases in brightness, the fields of which, in terms of the reflection (or transmission) coefficient, do not go to arithmetic progression(10%, 20%, 30% ...), but differ from each other exponentially (1%, 2%, 4%, 8% ...) - and this is nothing more than a logarithmic dependence. You probably know that the musical row, its frequencies (the vibrations of the string) differ from each other also exponentially. The same can be said for the power of sound, which is measured in decibels you know.

So, the human eye perceives the ratio of shades according to the logarithmic law, therefore, in scanning techniques, etc. it is this scale that is used. A change by D = 0.3 upwards indicates that the eye sees an object 2 times darker. The density is measured in bels.

D max - maximum density; D min - minimum density; ΔD - the ratio of certain densities, usually D max -D min; ΔD scanner - the range of densities (D max -D min) that the scanner can perceive.

How the research was conducted

D In order to have a wide range of densities, I used a sensitogram of black-and-white photographic film, I know all the absolute densities of its fields (taking into account the minimum density D min, or, more simply, taking into account the "veil density"), measured by the "M" status densitometer. Scanning b / w negative, as a rule, occurs in a "mixed" channel, so I will scan this one. For a field with a density D = 0.0, I took the luminescence of the lamp itself, i.e. the scanned area of ​​the image without film. The sensitogram had a maximum blackening D max = 2.3, in order to obtain a blackening with a density D max = 2.6 I used a neutral gray filter with a density D = 0.3, pressed against the area of ​​maximum blackening of the sensitogram. Scanning was performed with the Xsane program (Linux platform) at a resolution of 300 dpi in b / w mode without any adjustments (brightness, contrast, gray level), the Xsane's ability to set the brightness with hardware was not used. The resulting 16-bit file was measured with a 5 × 5 pixel "pipette" in Photoshop.

Results:

D test	0,0	0,3	0,35	0,4	0,48	0,54	0,65	0,8	0,9	1,0	1,15	1,3	1,4	1,5	1,6	1,7	1,8	1,9	1,96	2,06	2,1	2,2	2,3	2,36	2,4	2,5	2,6
D scan	0,0	0,17	0,2	0,22	0,26	0,3	0,36	0,43	0,5	0,57	0,63	0,72	0,8	0,85	0,92	0,96	1,1	1,1	1,15	1,15	1,2	1,3	1,3	1,4	1,4	1,4	1,4
%	0,00	33	38	41	46	51	57	63	68	73	77	81	84	86	88	89	92	92	93	93	94	95	95	96	96	96	96

Where: D test - density in the tested negative;

D scan - the value calculated from the Photoshop blackening percentage "a to white;

% is the percentage of blackening measured by Photoshop.

It is rather difficult to analyze the obtained values ​​without preparation, and it is not necessary. Based on these data, a graph (characteristic curve) was plotted, the D test values ​​were plotted on the X-axis, and the D scanner values ​​on the Y-axis.

Analysis of the received data

T Now it is much easier to analyze the graph :-) So, what we see: the graph curve up to D test = 1.6 is quite flat and smooth (marked in green), which means the scanner transmits values ​​up to this density almost proportionally, without distortion.

Between D test = 1.6 and D test = 2.35, the curve looks like a broken line (marked in yellow), so I dare to assume that in this part of the characteristic curve the scanner produces "well thought out values". Those. the matrix perceives them, but gives out something incomprehensible, in order to "digest" them into a "normal" form, the scanner has to correct these values. This can be compared to the "decibel" in professional video cameras. When the illumination level of the object is not enough, the operator turns on the "decibel", the camera begins to increase the level of the signal received from the matrix, in fact, the electrical signal is amplified. Both what is needed and what is not needed increases. Thus, simultaneously with the image, noise is amplified. Something similar happens in the scanner: noise appears in this section D test, so the curve looks like a broken line.

And now for the fun part. Who wrote there about ΔD scanner = 3.0 for this scanner? Well, well ... For a value of D test = 2.35, this scanner does not perceive anything at all! So ΔD epson_perfection_1650_photo = 2.4!, and even then, only because D test = 2.35 is the last field that has a different value returned by the scanner from the previous one. You understand, except in red, I could not highlight it :-)

Outcomes:

• The scanner is capable of normally, almost without distortion, perceive the density of a transparent original up to 1.6;
• The scanner, while introducing distortions and "noise", is still capable of detecting densities from 1.6 to 2.35;
• The scanner is blind to the density of 2.4, any density above this value it perceives as black.

What to do?

D Let's see what the scanner manufacturer has to offer. In Xsane (to be precise, in the backend "e Sane) there is an opportunity to adjust the brightness with the help of" iron. "That is, the scanner, as it were, increases the brightness of the lamp in order to" break through "D max = 2.4. in fact, there is no increase in the brightness of the lamp, the scanner (or rather its firmware) processes the obtained values, as a result we should get a higher value of the maximum density, which the scanner interprets as black. So, we will use the capabilities provided by the manufacturer. Set the Brightness value in Xsane to the maximum, which allows “iron.” In our case, it is 3.

As in the previous test, we build a graph based on the results obtained (so as not to overload the reader with information, I do not give them).

For comparison, the first characteristic curve (test 1) was left, the new curve (Brightness = 3) is marked in red (test 2). Let's proceed to a comparative analysis: the scanner, as it had ΔD scanner = 2.4 and has, on the basis of which it can be judged that the "decibel" (signal amplification mode) is always on, and works in the section D test = 1.6 D test = 2.4, since the scanner cannot distinguish any new, higher D max_test values.

The characteristic broken line in the section D test = 1.6-2.4 has become smooth, which indicates that the firmware of the scanner, when the brightness increase option is enabled, converts the values ​​obtained from the matrix more correctly from the point of view of tone rendition. But judging by the images, this does not make the "noise" less, it only becomes more, as it intensifies, or, perhaps, the "noise" becomes smoother. The latter is most likely true.

Now let's look at the section from D test = 0.0 to D test = 0.5, the curve in this section has a low gamma value. That is, the lights will be transmitted softly, and lighter than they actually are.

Let us evaluate the result obtained as a whole: the increase in brightness occurs not due to the effective use of densities, but due to the change in the level of all densities (note which tone is used to convey the "black" value, if in test 1 it is at the value D scanner = 1.4, then in test 2 on the value of D scanner = 1,2). There is no point in using this option. We will not get any useful increase in brightness. The “gray field” will become lighter; The "white field" will remain the same as it was; The “black field” will also become lighter, but no new details will appear there. The scanner “saw” D scanner = 2.4 and “sees”. But the level of "noise" will increase.

To be honest, when I did this test, I thought that Epson would still "shift" the curve to the right, i.e. we will lose details in highlights, but get in shadows, i.e. D scanner will not change, but will work in a different section D test = (D max -D min). Perhaps the manufacturer tried to implement this feature. This is indicated by the characteristic curve in the range D test 0.0-0.5. I will assume that this was done in order not to lose detail in highlights if the curve is shifted to the right. In practice, only the average gradient has decreased.

Scanning black and white negatives.

NS we try to prove the results obtained in practice. For the "purity" of the experiment, I will use one single black and white negative all the time. Note that the negative used has normal densities and is developed to an average gradient of 0.62, which is de facto the standard. In the film laboratory, it is printed at 11 light, which is the norm.

As we have already found out, one of the problems with scanning both negatives and slides is the presence of "noise" in the image. This phenomenon is especially noticeable when scanning sufficiently thick (dark) originals. This is due to the limited range of optical densities ΔD scannner = D max -D min.

For example: the Nikon Coolscan 4000 scanner is able to reproduce the optical density range of 4.2 (I don't want to upset anyone ... about the Epson 1650, I have already found out its ΔD = 3.0 :-)). Simpler scanners have more modest performance.

The maximum range of optical densities of a b / w negative is 2.5, ΔD max of a slide = 3.0, a color masked negative is about 2.5, but due to the presence of a mask, this type of negatives has a large D min.

I am convinced that a ΔD scanner = 3.0 is sufficient for scanning anything other than, perhaps, X-rays. The problem is in which part of the negative (slide) is this ΔD scanner = 3.0. I'll try to explain why.

Let's discard the knowledge about photographic paper, it can be high-contrast, contrast, normal, semi-soft, soft. We will use normal paper in this example, because adjusting contrast with positive material is a "crime". The positive should be standard (these are the rules in cinematography, and in darkrooms too), the contrast should be decreased / increased - work with the negative (change the development time, do remote sensing, use filters, countertypes, etc.). So, let's use the standard positive.

Do you know what range of densities the positive is able to reproduce? ΔD = 1.0! Total!

The data are given without considering the minimum density.

That's it! Thus, photo paper does not reproduce the entire range of negative densities, it is not necessary, it is harmful! The result is a wildly “soft”, low-contrast, “not juicy” picture, even if the print contains both white and black fields! If you don't believe it, find the negative with such an interval (ΔD = 2.5) and scan it! It is still a problem to find it ... Here I used a sensitometric wedge (the same one), I know its density: black field (veil) - 0.3; white field (maximum blackening) - 2.3, thus ΔD neg = 2.0. A point with a density of 0.3 was assigned “black”, a point with a density of 2.3 was assigned “white”, then in the same mode I scanned a sample of our negative. "Beauty", right? I must admit that I raised the gray level a little, the negative turned out to be completely dark. But the critical points of black and white remained in place. So the average gradient hasn't changed.

Further, in accordance with the sensitogram, for a field with a density of 0.1 (above the veil) a "black point", a field with a density of 1.1 - a "white point", and for a field with a density of 0.6 I assigned a "gray point", those. I simulated normal photographic paper. Here's what happened:

What conclusion can be drawn from all of the above - yes, the fact that the negative contains a huge amount of densities that are not printed in the positive. At the beginning of the 20th century, there was a tale that the average gradient (contrast ratio) of the negative, when multiplied by the average gradient of the positive, should give 1.0, then, they say, the gradations will be conveyed in the correct tones. What is the bottom line? - sluggish images! The piece should be 1.7 ~ 2.2.

Thus, even ΔD scanner = 1.7 is enough to scan a negative in case we want to simulate "extra soft" paper.

For clarity, I have marked a useful interval of negative densities on the characteristic curve graph. A test object with such densities (a pretty girl and a number of gray densities) is supplied by film manufacturers for tuning minilabs.

As you can see, the useful range of negative densities fits into the “safe” range of densities perceived by the scanner without any difficulty. If we have correctly exposed the film, then we can even afford D min = 0.5, but for b / w negative (not masked) this is a very high minimum density.

What conclusion can be drawn? For scanning a normal b / w negative, ΔD scanner = 1.6 ~ 1.7 is more than enough.

Scanning color masked negatives

TO As mentioned above, the color masked negative has ΔD max = 2.5, while having high values ​​of the minimum density D min. For example, the Fuji color negative I measured had the following D min values:

If you think roughly, then this is almost the norm (there is no GOST at hand). Now we add the values ​​of the useful density range of the color negative (they are the same as for b / w film) with the D min values ​​for each channel.

For clarity, we note this on our characteristic curve graph (the characteristic curves of all three channels are similar; it is quite acceptable to depict one)

It's not hard to see that Red the channel fits into the “safe” zone without any problems, there is even a small margin; green the channel enters the "dangerous" zone with dark areas of the negative (in the positive they will become lights); blue the channel enters the "dangerous" zone by half, from gray to white in the positive.

Therefore, in red there will be no "noise" on the channel; v green on the channel, “noises” will appear in the bright areas of the positive; v blue on the channel, "noises" will be from gray to white. Let's try to confirm this.

As I said, I will use the same black and white negative. To simulate color masking film, an unexposed section of Fuji color negative film was superimposed on the negative. I will also show the histograms of the results. So, let's scan the “color” negative!

Due to the presence of an orange mask that turned around and turned blue, the positive looks blue. We don’t want to see it blue, what should we do? Increase the gamut of the blue layer with the "software" so that the "white" field becomes not blue, but white. Well, let's try. Move the "sliders" on the histogram so that the image becomes neutral gray in all densities, from black to white.

And, lo and behold! The picture is normal in color, well, almost :-). Now let's open it in a graphics editor and take a look at the image disassembled by channels:

Red

Green

Blue

There is almost no noise in the red channel, in the green channel they are not large and quite acceptable, but in the blue channel there is a lot of noise. This is not scanner noise, this is a problem of scanning masked films, or rather, the "stretching" of the blue channel. To prove this, I scanned the same black and white negative, but without a mask in RGB mode and I will also demonstrate it in the form disassembled into channels:

Red

Green

Blue

As you can see, there is no noise in any of the channels. So, our "enemy number 1" is a yellow-orange mask! More precisely, the high minimum density behind the blue filter. And you have to fight with it.

Of course, these problems do not arise during photo printing, photographic paper (not Soviet :-)) is already balanced in terms of the photosensitivity of the layers to match the orange color of the mask. Modern colored photographic papers have a light sensitivity to blue rays about 20-30 times higher than to red ones. The fact is that photographic paper (in photographic enlargers, in photo printers) is exposed not with white light, but with the yellowish light of an incandescent lamp, moreover, having passed through an orange mask. In scanners that are not specifically designed for scanning negatives, matrices are balanced to digitize slides and non-masked negatives.

Scanner manufacturers are trying to solve this problem in a number of ways. My Epson, for example, allows you to scan a 48-bit image, at 16 bits per channel, so there is something to "stretch". There is, of course, an effect. Compared to an 8-bit picture, the difference is colossal. Nikon, in its scanners, uses an expensive matrix capable of "seeing" ΔD = 4.2, but there are other problems because of this :-)

By the way, not only color negatives are poorly scanned on Epson, but also dense (admittedly dense, of course) b / w negatives, as well as dense slides. See above for the reasons.

Thus, what is preferable for photo printing (overexposing the negative by ½ aperture) becomes a scanning disaster. How to deal with this? What to do?

What to do? Take two!

T Oh, what about photo printing: increase the exposure!

If during photographic printing we can increase the shutter speed or slightly open the aperture, then during scanning we can only increase the brightness of the light source (i.e. the lamp). Although, in the version "from the manufacturer" we cannot even do this. I, at least, have not heard about the implementation of this opportunity in "budget" models. This is all great, of course, but only applies to scanning black and white negatives. In the color version, it is necessary to use the exposure control in three channels (in fact, two are enough - for the blue and green channels, I have never seen a blue mask). There are different ways to do this:

1. Use a color mixing head from a color photomagnifier, or color filters opposite to the mask color (for example, a blue compensation filter for incandescent lamps) to "neutralize" the mask, so to speak, to make it a neutral gray. And to increase the brightness of the lamp, in order to "break through" the resulting equal D min_negative across the channels.
2. Use three passes (one per channel) with different exposure for each channel.
3. Solutions for manufacturers:
   • use lamps different types for scanning color negatives (with a higher color temperature), and slides;
   • use lamps of higher brightness (with a margin), and the ability to reduce this brightness (it seems a good idea to use a gray filter inserted in front of the lamp, no changes in color temperature!).
   • Use two matrices. One balanced for slides, one for masked negatives (expensive way).

What should an ordinary user do? I think that the solutions described in the first and second paragraphs can be implemented at home. The first option seems more realistic to me. At least, you can make a Preview without using specific "software" (no one wants to write? :-)). For example, make a "light box" with the ability to insert filters and thus adjust the color and brightness of the luminous flux. Or use a colored head from a magnifier. And leave the native lamp for b / w negatives of the density normal for the scanner, as well as normal slides.

Still, why is ΔD scanner = 3.0 enough

D but because if there is a high density on the slide, then most likely it is not needed, you need to be able to use at least ΔD scanner = 3.0, but in the place of the original density interval where it is really required. The problem is in which part of the negative (slide) is this ΔD scanner = 3.0. There is simply no point in making the ΔD scanner larger, and in the case of Coolscan, it is even harmful. Because as a result, a rather soft (or low-contrast) picture is obtained from the negative. Any increase in contrast, or gamma, with the help of software increases the level "Noise." True, you can scan an image with a resolution of 4000 dpi, make all the adjustments, and reduce the resolution. But then it turns out that 4000 dpi is needed only in order to then reduce it, suppress the noise? It turned out confusing ... sorry. In case, this is a very good scanner, for the money it costs. In short, you need to increase not the ΔD scanner, but add the ability to adjust the exposure!

Take your negativity back! I need a slide!

TO Once, I didn’t have a very good idea why printers hate to scan negative, there were a lot of assumptions: they don’t want to mess with the color rendering, raise the contrast - and all that sort of thing. The main reason is different. In principle, there is always "noise", either they are visible or not. So, from all of the above, it follows that "noises" tend to appear in the darkest parts of the original. When scanning a slide, the "noises" appear in the shadows, and it is quite problematic to see the "noises" in the shadows. When scanning a negative, "noises" also appear in its darkest areas. And everything would be nice if it was not necessary to reverse the negative. Have you already guessed? When negative turns into positive, “noises” appear in the lights, and it is not difficult to examine them, but not noticing is really a problem. In addition, with modern implementations of scanners, even professional ones, it is practically impossible to scan a negative with high quality! To do this, you need to control the exposure. Do you know such scanners? If yes, please email me the names and, if possible, links.

What about the new b / w masked films?

WITH I think that Leonid Vasilyevich Konovalov made this “new” film on “Svema” back in 1989 (I can lie, but those times), in order to “painlessly” use b / w film frames in color printing. Well, okay ... The main color of the mask is "orange", therefore, red rays pass through it best. As a consequence, the mask has the lowest minimum density in the red channel. Just scan the red channel. If there is no such option in your driver, scan RGB and take the red channel; "The rest" can be thrown out :-).

What does a housewife need?

D In order to scan a standard negative with high quality, a housewife needs a scanner with a ΔD scanner> = 1.7 and three "pens". Two for adjusting the amount of blue and green light, and a "knob" for adjusting the overall brightness of the light source. To scan a standard slide, you need a scanner with a ΔD scanner> = 2.5 and a "knob" for adjusting the brightness of the lamp.

Conclusions:

1. The Epson Perfection 1650 photo scanner has a ΔD scanner = 2.4, a useful density range of ΔD scanner = 1.6.
2. In the form in which the scanner is supplied by the manufacturer, it is suitable for scanning:
   • b / w negatives, including masked ones ( Red channel);
   • slides of normal density with small amount dark areas;
   • unmasked color negatives (remember the Soviet film DS-4?);
   • the scanner is conditionally suitable for scanning color masked negatives (the practical use of these "scans" is questionable; they are suitable only for "previews").
3. The denser the original we scan, the more "noise" we have.
4. The scanner can be adapted to scan color masked negatives by attaching a higher power lamp to it and using color (blue-cyan) filters to correct the color of the light flux.

Lyrical digression (cynical)

V in general, this is a normal situation when scanners are made by people who, apart from the photograph of their wife, have not scanned anything and have little knowledge about negatives, positives, and the rest of the "nonsense". Movie cameras (and not only cameras) are made by people who don't work in movies. The same guys (a stone in the kodak and fuji garden) came up with a mask for a color film (if someone does not know, there is little use from it, practically no) and a fourth violet-sensitive layer, instead of changing the spectral sensitivity of the red layer. It is because of these “friends” in our country that instead of our normal one, the wrong standard for measuring densities was introduced (but corresponds to the world!) ... So, a lyrical digression ...

How do you scan?

F The original Epsonian "firewood" is only suitable for checking the scanner's performance upon purchase, and for scanning texts (in 48-bit mode :-)). I am using Linux with the Xsane program because there is a "carriage and a small cart" manual settings, including hardware settings. And most importantly - Xsane is worthless! Why don’t I use SilverFast ?, because I don’t have it :-), and my demo version "made me live a long time." If someone gives - I will not be offended :-). One of these days I'll try VueScan, they say it's a good scanning program, and there is a version for Linux. There are plans to screw the color head from "Krokus GFA" to your scanner. I think I will do it in the near future.

In the photo there is a landscape not far from the Kazanskaya village of the Rostov region.

Gratitude.

V I express huge gratitude Leonid Vasilievich Konovalov for his help in correcting, in his own words, "spelling" errors.

Materials used in writing the article:

• L.V. Konovalov, "How to understand the films", VGIK, 1997.
• V.A. Yastold-Govorko "Printing photographs", "Art", 1967.
• Materials of the site bog.pp.ru

A responsibility?

A what it is? :-)

The author's opinion on the above issue is not "the ultimate truth." I only state what I checked, tried, "felt" ... The opinions, conclusions, results and statements of the author may not coincide with yours, or someone else's. The recommendations given in the article should not be taken as a guide to action. All suggestions that you may implement in your equipment after reading this article, you make at your own peril and risk. The author does not assume responsibility for any damage that may be directly or indirectly caused by using the recommendations set out in this article.

Copyright

NS This article, as well as its translations, may be reproduced and distributed in whole or in part in any medium, physical or electronic, provided this copyright notice is preserved on all copies. Commercial distribution is permitted and encouraged; but the author of the article would like to know about such use.

All translations and derivative works based on this article must be accompanied by this copyright notice. This is to prevent restricting the free distribution of this article. Exceptions can be made in cases of obtaining special permission from the author, who can be contacted at the address below.

The author would like to distribute this information through various channels, but at the same time retain copyright and be notified of all plans to distribute the article. If you have any questions, please contact the author of this article on e-mail: [email protected].

Vasily Gladky, 2003

Don't be surprised if you don't find these words in the specs of your scanner - manufacturers don't always quote this metric. But this does not mean at all that this characteristic does not play a significant role in the quality of the resulting image. On the contrary, many experts agree that this is the main indicator of the quality of the scanner.

What is dynamic range?

More precisely, this parameter is called the optical density range.

Optical density is an indicator that allows you to numerically measure how dark the original is. For a transparent original, optical density is the decimal logarithm of the ratio of the total light flux to the light flux passing through the original; for opaque ones, the ratio of the entire stream to reflected light.

Thus, the darker the original, the higher its optical density. For example, a value of optical density of 0.01 corresponds to almost white light, and values ​​of 4.0 and above - to almost black, almost invisible to the eye.

On any slide, there are both light and dark areas - a whole range of different optical densities. The range between the lowest and highest optical density on a given original is called its dynamic range.

Scanner dynamic range

Not only the original has dynamic range, but also the scanner. The dynamic range of a scanner is the difference in optical density that the scanner can recognize.

All scanners recognize white color quite well. In other words, they have no problems with the minimum optical density. Most scanners have 0.01 or less. Problems arise when scanning dark areas where there is very little light. It all depends on the sensitivity of the reading photocell: the more sensitive the CCD ruler, the better the scanner recognizes dark areas.

What does "recognize" mean?

This word means two actions at once. First, the scanner must distinguish between a dark shade and a maximum black shade. Otherwise, many of the dark areas in the scanned image will appear as just a black spot without any detail. Secondly, the scanner must scan a dark area without noise - a kind of colored debris in the form of multi-colored dots. After all, the darker the original, the weaker the signal on the photocell, and the greater the contribution to the image will be made by the noise of the photocell itself and other electronic components of the scanner.

The scanner's ability to distinguish between dark and black areas and the amount of noise in dark areas are usually related. They are determined by the quality of the photocell and the color depth of the scanner: the darker areas the scanner detects, the less noise the photocell introduces.

Therefore, these two parameters are usually combined by one characteristic - the dynamic range, which shows how high-quality the photocell is installed in the scanner, and therefore how much dark areas it recognizes and what level of noise in the shadows it gives when scanning. Of course, the greater the value dynamic range, all the better.

In addition, the dynamic range depends on the color depth of the scanner, that is, on the number of grayscale (brightness) that it can transmit. This is natural: the fewer brightness gradations the scanner transmits, the less the difference between the lightest and darkest shades that it recognizes.

These parameters are linked very simply. Let's say the color depth of the scanner is 36 bits, or 12 bits per color. This means that it recognizes 4096 shades of gray. Decimal logarithm of 4096 gives 3.6 - this is the maximum dynamic range of this scanner. In fact, it is less, since the sensitivity of the photocell is not ideal. How much depends on the quality of the photocell. However, we can say for sure that the dynamic range of this scanner cannot exceed 3.6.

The dynamic range can be used to accurately classify scanners (Table 1).

Dynamic range of original

Obviously, the dynamic range value of the scanner must be greater than the dynamic range value of the original. Otherwise, when scanning, some of the information from the original will be lost: if the image is not completely black, then the dark shades will disappear. For example, instead of a shadow on the face, there will be just a black spot. Or the scanner will raise the brightness of the image and recognize the dark areas well, but instead of the light areas, you will get spots, this time - white.

Data for the most common opaque originals are shown in Table 2.

Therefore, the range of a scanner designed to scan exclusively opaque originals should be at least 2.3–2.5. On the other hand, it should not exceed these figures too much, since the price of the scanner increases exponentially with the increase in the dynamic range.

With transparent originals, the situation is somewhat more complicated. First, there are professional and amateur photographic materials. For the latter, the range of densities is somewhat less.

Second, unlike opaque originals, which are usually printed on white paper (that is, the dynamic range is measured from white with a low density), in negatives the lightest shade still has a significant density.

This means that when scanning negatives and slides, you must take into account not only the dynamic range, but also the maximum optical density. For example, a slide with a dynamic range of 3.0 might have densities between 0.7 and 3.7. But the dynamic range of the scanner is measured practically from white - from low densities. Thus, if the scanner has a range of 3.5, then the maximum density it can recognize is 3.55 (maximum is 3.6). Such a scanner will not be able to correctly scan the slide described above, although its dynamic range is higher than that of the original.

Therefore, for transparent originals, it is better to take into account not the dynamic range, but the maximum optical density (table 3). In other words, the maximum optical density of the slide must be less than the maximum density that the scanner recognizes.

How to scan?

Whatever the manufacturer says, the dynamic range of a flatbed scanner from the class "office and home", the so-called SOHO, costing up to $ 450, does not exceed 2.6-2.7. The CCD line alone, which can provide a dynamic range of 3.0, is more expensive.

This scanner handles opaque originals well, but dark areas on slides will look like a solid black spot with a lot of noise. If you try to scan a negative on such a scanner, then after inverting all the light areas (those that were dark on the negative), for example, a sky with clouds or a light shirt, will look like a solid white spot without any details, except for the same noise.

Therefore, even if you buy a slide module for the scanner for $ 200, you will not be able to scan slides and, moreover, negatives on it with high quality.

The minimum dynamic range, at which one can hope for a more or less decent result, is 3.0, and 3.4 is better. The minimum cost of a flatbed scanner with this range is $ 600. A slide scanner with 3.0D will not cost much less, and for professional use scanners with a range of 3.4D and higher are required.

What to scan?

We will not try to classify the originals, but just figure out which originals should be preferred and which ones should be avoided, and why.

Let's start with the simplest thing - scanning text. High resolution not required for this work, but subtleties are still there.

First, when choosing a scanning method, any scanner offers two options:

• mode black & white(halftone) - black and white without shades of gray;
• mode grayscale- with shades of gray.

In the first case, you can forget about the drawings. They will turn into black spots, leaving only the text. Moreover, if the text is not very clear, in places blurry or just pale, then the resulting image will look pitiable.

On the other hand, black & white is the fastest and most economical in terms of file size. It should only be used for very clear text.

Otherwise, grayscale scanning is best. A text recognition program will perfectly cope with such a file, and drawings, logos, etc. will be scanned normally.

If the original is in color, consider the capabilities of the scanner.
In principle, the best original is a slide, a little worse is a negative, even worse is a photograph, and printed color prints like clippings from magazines are generally better avoided.

Why?

First, it is in this order that the dynamic range of the originals is reduced. But this is not the main reason why slide or negative is preferable to photography.

The fact is that each original is characterized by a color gamut - a set of transmitted shades. This parameter should not be confused with color depth. Color depth shows the number of shades, and color gamut shows which shades they are.

Let us illustrate this with an example. The human eye has the largest color gamut. It can be depicted in the form of a certain figure, which reflects all the perceived shades (see figure).

The large triangle outlines all the shades that the slide and film in general convey, the smaller triangle corresponds to the colors transmitted by the monitor (the outline for the scanner is something in between the slide and the monitor). Finally, the inner shape corresponds to the CMYK ink set, that is, the color gamut of the printing machine (and color laser printer, which has a slightly larger color gamut).

Thus, the green-blue gamut is well transmitted by the film and the scanner, but not by the printer (it is a known fact: on a standard 4-color printer, you cannot depict a blue sky).

Hence the moral - if there is a choice, then you need to scan the original that transmits large quantity shades, that is, a slide, and not a photograph printed from it. However, not all scanners can scan slides because of the weak dynamic range of office models. Therefore, the owner of a $ 100-200 scanner often has no choice.

It is necessary to say separately about printed prints. Printers and printing machines print with special dots - a raster, the frequency of which is not too different from the resolution of the scanner 1. Do you want to know what happens if you superimpose two periodic structures - a scanner and a print - on top of each other? Look at the light through two layers of nylon or any other translucent synthetic fabric. You will see moire. The same moire pattern will result from scanning a printed print.

The special Descreen function in the scanner driver helps to combat this effect. It removes moire by slightly blurring the image. But at the same time, quality suffers significantly. Therefore, you can scan clippings from a magazine only with the subsequent reduction of the image, then the blur effect will not be so noticeable.

Brief summary - if the scanner allows, scan slides, not photographs. If possible, avoid scanning printed prints, and if there is no way out, then scan with the subsequent reduction of the picture, at least 1.5 times.

Original view... Scanning can be done in transmitted light (for originals on a transparent substrate) or reflected light (for originals on an opaque substrate). Scanning negatives is particularly difficult because the process is not limited to simply inverting the gradations of color from negative to positive. To accurately digitize color in the negatives, the scanner must compensate for the color photographic haze on the original. There are several ways to solve this problem: hardware processing, software algorithms for moving from negative to positive, or look-up tables for specific types of film.

Optical resolution. The scanner does not take the whole image, but line by line. A strip of photosensitive elements moves along the vertical of the flatbed scanner and takes an image line by line. The more light-sensitive elements a scanner has, the more dots it can remove from each horizontal stripes Images. This is called optical resolution. It is usually counted by the number of dots per inch - dpi (dots per inch). Today, a resolution level of at least 600 dpi is considered the norm.

The speed of work. Unlike printers, the speed of scanners is rarely specified, since it depends on many factors. Sometimes the speed of scanning one line is indicated in milliseconds.

Color depth measured by the number of shades that the device is able to recognize. 24 bits correspond to 16,777,216 shades. Modern scanners are produced with a color depth of 24, 30, 36, 48 bits.

Dynamic range characterizes what range of optical densities of the original the scanner can recognize without losing shades either in the highlights or in the shadows of the original. The maximum optical density of the scanner is the optical density of the original, which the scanner still distinguishes from complete darkness. All shades of the original darker than this border cannot be distinguished by the scanner.

Batch processing - this is to scan several originals at the same time, saving each image in separate file... Batch processing software allows unattended scanning of a specified number of originals, ensuring automatic switching scanning modes and saving scanned files.

Scaling range - this is the range of original zoom amounts that can be performed during scanning. It is related to the resolution of the scanner: the higher the value of the maximum optical resolution, the greater the magnification factor of the original image without loss of quality.

By interface type scanners are divided into just four categories:

Parallel or serial scanners connected to LPT or COM port These are the slowest interfaces. Problems may arise related to the conflict between the scanner and the LPT printer, if any.

USB scanners Cost a little more, but significantly faster. A computer with a USB port is required.

Scanners with SCSI interface, with their own interface card for the ISA or PCI bus, or connected to a standard SCSI controller. These scanners are faster and more expensive than the representatives of the two previous categories and belong to a higher class.

Scanners with modern FireWire (IEEE 1394) interface specially designed for graphics and video. Such models are presented on the market relatively recently.