Understanding Resolution - Do You Need More Pixels? in Photography

May 05, 2009, 06:44 PM
by Felisia A


http://Photography.dinogroups.com/dlink.cfm?blog_id=1561CS907Fg0Quc10aFEXXb0
Understanding Resolution - Do You Need More Pixels?

The limitation comes from the lens resolution. There''s a point where more megapixels simply becomes a burden without bringing any perceivable benefit at all. I personally believe with 35mm, it is probably around 20 megapixels. Of course, there is not a clear cut off line, but a gradual approximation. With 35mm full frame, the first critical point of megapixel arrives at 4M. I will show you why below. Before reaching that point, any increase of number of megapixel resulting a quite strict one-to-one increase of resolution. For example, 2M quite literally has twice as much much resolution as 1M. After that, this will start to degrade. For the matter of better communication, consider a coefficient called ratio between the perceivable resolution increase and the pixel number increase. This ratio is practically 1/1 before the number of pixels is 4M, but starts to fall off gradually after that. My estimate is that it would fall to less than 0.5 when the number of pixels reaches 10M, and probably less than 0.1 at 20M. I''m pretty certain that this ratio approaches 0 by 40M. What that means is that after 10M, the perceivable resolution increase one receives from every 1M increase of megapixels is less than .5M, and after 20M, the perceived resolution increase for every 1M is less than .1M. As a practical matter, after 20M, the more megapixels simply become "the Emperor''s new cloth" which really is not there to be seen. It just makes some people happy. And I''m talking about full frame 35mm. For APS sensors, the number is even smaller. All that is because the inherent limitation of the lens resolution. Take 35mm full frame. I''ll use approximation to simplify the explanation. Suppose there are N columns (or rows) of photosites for every 1mm. The total number of pixels of a full frame 35mm sensor is then (Nx36)x(Nx24), which is roughly 1000N^2 (pixels). Now, good 35mm lenses usually have excellent MTF at 30 line pairs per millimeter. Especially at the center of the frame, a good 35mm lens can have 80% or above MTF at 30 line pairs per millimeter. So it is reasonable to take 30 line pairs per millimeter as a baseline. A sensor that matches this resolution would have N=2x30 (two photosites for each line pair), and correspondingly the total number of pixels of 1000xN^2 = 1000x60^2, which is approximately 4M (or slightly less). This tells you two sides of the story. First, on a 35mm full frame sensor, before the number of pixels reaches 4M, most lenses have enough resolving power to take full advantage of the more pixels. Second, when the number of pixels goes beyond 4M, most lenses will not be able to fully take advantage of the increase of the pixels. That of course doesn''t mean that after 4M, the increase of the pixel numbers will have no benefit at all. They still do, but will not have the full benefit due to the lens resolution limitation. For example, suppose you have a lens that has an MTF value of 50% (or MTF50) at 50 line pairs/mm (meaning that at this resolution, the lens delivers a perceivable contrast of only 50% of the full contrast, i.e., losing 50% of contrast). A sensor that matches this resolution would have N=2x50=100, and correspondingly a total number of pixels of 1000xN^2 = 1000x100^2 = 10M. That means, with this lens, a 10M sensor is able to represent each line pair with two columns/rows of photosites, but with a loss of 50% of contrast at such resolution. It could be said that at this level, you are only getting 50% of the perceivable resolution out of the increased number of pixels. Now, those who are familiar with lens tests will tell you that a 35mm lens with an MFT50 at 50 line pairs/mm is a very good lens. For 35mm full frame, this is about 1200 lp/picture height. With a 20M sensor, you need a lens that has an MTF50 with 70 lp/mm or an equivalent of 1700 lp/picture height for full frame. This is already surpassing the optimal resolution (center of the frame with a stopped down aperture) of even the best prime lenses. And we are talking about MTF with a 50% loss of contrast here. Overall, even with a full frame, going from 10M to 20M has less than 50% increase in perceivable resolved pixels even with the best lenses. You''re certainly not doubling the resolved pixels there. Of course, even an increase of 50% can be important for some people who (1) actually need a greater resolution; and (2) actually have possession to the top lenses and use them. On the other hand, with the same technology, going from 10M to 20M will lose at least one full stop of the ISO performance. I have no problem with people''s preferences. I just have a problem with people making choices without the necessary information and understanding. Still, one could make a rational (albeit discounted) argument for 20M for a full frame sensor. But beyond that, it certainly looks like an Emperor''s new clothes to me, unless they suddenly come up with drastically better optics, which is extremely unlikely. Besides, unless you want to make prints larger than 30 inches, you really don''t benefit from more than 20M anyway even if you had a lens that could resolve more. And you''re not getting those large files as a freebie. They impose a heavy tax on your computer system.






Dinogroups™ is powered by VEELABS. Please also visit
DinoMarket - Tempat jual beli, pasang iklan dan buat toko online gratis.
DinoBar - The stylish online shopping for gifts and flowers