Theme: “What is better: the figure or film?” – One of the most talked about on the Internet. Discussions will continue for a long time, anyway, so long as the purely economic reasons, digital technology finally prevail.
After the publication of the article “How many pixels per frame narrow screen the film” I got quite a lot of questions and interesting comments. In addition, I had to discuss with an issue in some Internet forums. As a result of this communication and the real answers to frequently asked questions.
I do not claim to be the ultimate truth. I want just as much as possible to clarify its position. Any comments on the merits issued, as always, welcome.
1. Question: “What these calculations? Do you believe that art can be measured and expressed in numbers? ”
A: Did I mention art? I answer the private technical issue. No more. For art, of course, the number of some sort of pixels does not matter. Fine, you can create and work to a large paintbrushes on scrim. No doubt I have on this account no.
2. Question: “Are you sure that good quality images of 20 million pixels is enough?”
A: About the quality, it also does not go. I used the criterion of equivalent resolution to answer the question posed in the title. Quality – the notion of a more complex and partly subjective (at least until we introduce a formal criterion that someone will arrange, and someone there). Talking about quality, we are always talking about something subjective. To understand my position on this issue and help others posted here answers.
By the way, there is no sense in the discussion “What is the best figure or film?”. Everyone chooses what suits him, and the fact that most meets their needs and requirements.
3. Q: “Explain more in detail as to why permission should be three pairs of lines in pixel?”
A: These are some of the fundamental constants, which are used in various fields. Back in 1933, Kell (Kell) found the value of this constant experimental way. Together with a group of other engineers, Kell studied at the resolution of the broken horizontal lines in black and white images. He found that to determine the resolving power should be the number of rows multiplied by 0,7. This correction factor in television is called a factor Kell (Kell-factor). However, 0,7 – is about 2 / 3, ie, Kell found that to resolve a pair of lines needed three rows.
With this constant can be found in the printmaking process. Publishers require that the images were provided with a resolution (in dpi) and a half times (3 / 2) more than typographical lineatura (in lpi).
Also explained why the two pixels to resolve a pair of lines is not enough. If the lines are arranged so that one half (the thickness), black (white) line is one pixel, and the second half – on the other, then obviously the line will not allow. The coefficient 3 / 2 is precisely takes into account randomness and error location in the space permitted lines and recording digital matrix.
Sometimes you can find the view that the coefficient 3 / 2 is associated with the mosaic structure of the digital matrix. It is not. Kell got this rate for black and white image.
4. Question: “Some digital matrix now provides a resolution of 50 pairs of lines per mm. Perhaps, the film is already dead? ”
Answer: No, still alive. The point is that the current matrix, which provide such authorization, the size smaller than standard narrow screen frame. Therefore, to obtain equal-sized paper prints from digital image matrices have to be increased in a greater number of times. The matrix should not only give permission to 50 pairs of lines per mm, but also be of no less than 24 x 36 mm. Let’s not rush things.
5. Question: “It seems that you are comparing a system of” film + lens “with a digital matrix without the lens? Is this right? ”
Answer: Good question. The text does not really make clear this point. Of course, to be compared (and it is this I call!) System “film + lens” and “Matrix + lens”. In my comparison of the text is like a system of “matrix + perfect lens”. I answer the question: what is the most minimal number of pixels is not the matrix, so that it can provide the same resolution, that film. Given the number of pixels optics must be larger. But for the initial evaluation of the number of pixels is good enough. Details not worth discussing, so as not to overload the reader extra physical details.
6. Full criticized my article, Sergei F. With the kind permission of the author, is the full text of his critical notes (spelling and style of the author is fully preserved):
“I do not understand why you even knowing Nyquist theorem, make common mistake – the number of scanned pixels equates to the number of elements in the film? And then you refer to Sony, which believes that to save (when photographing or scanning) into 2 lines in the original image must take at least 3 pixels (camera, scanner). Ie Knowingly scan should be 1,5 x1, 5 = 2.25 times more than the original, with an equal volume of useful information … Strictly speaking, the information there is still a bit less to extract it more fully, it is better to crawl than with sesquialteral, but with double, and with even greater margin in frequency. What is done sometimes with the help of drum scanners with its 8000dpi. Only the size of the file in this case grows enormously, and vyiigrysh in the amount of additional information – a small, 15-20 percent.
If you go back to the film and count the number of “original pixels”, it contains the belief that a necessary and sufficient is scanned at 4000dpi, then get a maximum of 2145321: 2,25 = 953476piks = 9.5 megapixel. This is, generally speaking, very much. If this were so, then print 30×45 would look perfectly smooth, almost the same as 10×15. Alas, it is not so – the resolution of the film is very uneven across the square frame – sel ‘avi. It dramatically decreases with the lack of exposure – in the shadows (in the twilight, in chetverttenyah) are at a real shot a lot of … As a result, already at a print 10×15, made even with the Real, which is sharpest and the best (no kidding), sometimes comes out the opposite grain (“picture element “)… It is easy to figure out how little these places permit. Good figure it suffers to a much lesser extent (at 100-200 ISO, of course) – noise in the shadows are very small, to the same noise pixel – domain (grain) is not formed.
At present the figures, of course, also has its disadvantages, reducing informative. First of all, Bayer tricolor filter, use anti-aliasing filter.
But as the outcome, I think it is fair to conclude by Norman Koren, who deduced that the resolution of 35mm Provia frame corresponds to 11 megapixel Bayer matrix, but D1s – Image quality will approach the 645 format – (c). What’s real and watched. ”
It says a lot. Therefore, the answers will be few.
a). Yes, I know Kotelnikov-Nyquist theorem, and even know how to take the convolution of two functions. 
This assessment (9,5 – 11 megapixels) is very common, but erroneous. However, give the reader to decide who is right. Do not multiply entities beyond necessity. To answer the question about the number of pixels in a film frame is not needed nor Nyquist theorem, no discussion about scanning. I think Sergei embarrassed the first version of my reasoning, I offer for absolute amateurs. But for people at least a little prepared, I brought in his article approach № 2. I propose once again to trace the course of my thoughts:
Filmed. I do not crawl. We consider the film under the microscope and find that the resolving power of 50 pairs of lines per mm. Then ask yourself, how many pixels we will need the full-sized matrix to provide the same resolution, provided that the optics are so good, that permission does not ruin? Remembering known factor Kella, answer: at least 20 million
The question is: where do scanning, Kotelnikov, Nyquist, etc.?
Next: indeed, the resolution in the center and the edges of the film is not the same. This thought leads to think about reducing the assessment of 20 million pixels. But on the other hand, 3 pixels for a couple of lines – the minimum score. And Sergei admits it. And the difference between real and ideal lens must be taken into account. That is, an assessment of 20 million pixels and can be increased. We assume that these factors balance each other and stay on the assessment of 20 million
b..) Sergey passing mention of information. About information and Norman Koren wrote in the seventh part of his article. But everything connected with the information – a separate issue, unrelated to the number of pixels directly. The number of pixels rather it is directly connected with the resolution. The quantity of information in the frame and film, and digital is much less than it gives directly the number of pixels. This allows us to painlessly (without visible artifacts) uzhimat 8-10 times almost any subject in the jpeg-file. This allows us to use the mosaic matrix in digital cameras. Etc. In this context, it is better not to operate on the term “amount of information”. This is a separate big topic.
in.) Sergey wrote “… in some places the grain comes out the opposite (« picture element ») …»
Suffice porassmatrivat scans from film, to understand that the grain – not pixel. Grain is not entirely dark. In general, the film is much smaller grains pixel.
PM) A little bit of noise, casually referred to Sergei. The grain of the film, quantization of signals, different instability – all this leads to noise, which reduces the quality. Yes, it is. However, I suggest not to delve into such details in the popular articles. Moreover, the pro quality I’ve already written in response to the second question.
7. Question: “Exploring the properties of the darkening of the crystals, we can conclude that the film could be part size, comparable to the wavelength of light. The image on film will always be different from the images on a digital matrix. Do not you think that talking about such small equivalent number of pixels as 20 million, really funny? Account must be conducted on gigapikseli! ”
Answer: It depends on the criterion. I used the criterion of equivalent resolution. I think the use of such criteria is justified in almost all cases.
Comparing the film without the lens with other systems has no meaning. It is not a direct manipulation of emulsion you are going to create her image? It seems to me, speaking about digital, and film photography, one should always take into account the optics!
Incidentally, I understand your position. With minor reservations. If you’re really important that lines and points, which can not be resolved, blended into a beautiful spot with smooth edges, and not unsightly blot, convenient for the majority, I fully accept your position. However, at such high demands You have to prepare solutions, ensuring their purity, to take care of incubation in the display, print their own photos manually, etc. And even small speck of dust should lead you to a state of shock. If you do not do all this, then your position is meaningless. By the way, you should also be irritating and expensive printing quality prints. I think they look disgusting to you. Your feelings, probably the same as a person with perfect pitch, heard a false note. I do not know whether it was appropriate to congratulate you on this or must quickly sympathize.