tsdevine
Offline
• • • • •
Upload & Sell: On
|
 p.2 #14 · R5 - 400 MEGAPIXELS Vs Sony A1 pixel shift | |
It comes down to exactly the right amount of resolution (if there were such a thing.) That's why I'm trying to separate the more objective "what benefit in resolution in possible" from the subjective "do I need that benefit in resolution (if it exists)." Whether anyone "needs" more resolution is a decision that is up to the individual. No one person can answer that question and say that it applies to everyone. But understanding what is possible seems to be an interesting discussion. Yes it's limited by the implementation of pixel shift, the sharpness of a lens, etc.
In terms of color resolution, in a traditional bayer demosaicing process, 2/3rds of the color information at each pixel is interpolated from nearby pixels. This interpolation can lead to color error at the pixel level, so sampling all colors at each pixel eliminates the interpolation and increases color resolution. The 4 frame pixel shift mode on Sony yields the same "size" file, but addresses that interpolation (as does the 16 but produces a larger image.)
Foveon tries to address that interpolation, by stacking the color pixels on top of each other. But I think the signal to noise ration drops across the colors, so it's not perfect either.
Anyway, I'm sure there are probably people here that can articulate this better than I.
gdanmitchell wrote:
Generally, there is nothing wrong with more resolution. It can't hurt, as long as other aspects of varying importance aren't unduly restricted. (While the benefits of 100MP for sport photography images might be pluses, the downside of slowing down burst rate might be a greater minus.)
We can imagine three options for the relationship between optical resolution and sensor resolution.
1. They are equal. This is basically impossible, since the resolution of a lens varies across the image field and according to aperture and potentially other things.
2. The lens could resolve more detail than the sensor. In a sense, this "wastes" the extra resolution of a higher MP sensor.
3. The sensor could out resolve the lens. This most accurately records the highest quality image that the lens can project on the sensor.
Super high resolution sensor or sensor shift systems put us in category three, which I don't think is a bad thing, as long as we don't have to sacrifice anything otherwise useful to get there. But it is worth considering whether the sacrifices needed to achieve it are worthwhile. For example, sacrificing burst rate may be worth it if you don't rely on burst in your shooting and you think you'll see a significant difference in your final photographs from the extra resolution.
I suspect there are some edge cases where sensor shift techniques may be worth it. But in many cases I have to wonder if the small (and only visible if we go looking for it very carefully) improvement in, say, a 30" x 40" print is worth having to deal with the additional steps when shooting and in post, plus the need to restrict subjects tot those that don't move, is worth it in many situations.
Above I saw the term "color resolution" used. I'm not sure what that means. I understand that pixel shift can register all of the image using all three color channels in a way that single exposures on RGB sensors cannot. I also understand that this theoretically means that each pixel generated by a blend of images using the R, G, and B shifted pixels could be more accurate. But are we actually having a real world problem with "color resolution" in our photographs from high resolution non-shift sensors? I'm not seeing it if we are.
YMMV.
- - -
Yes, it is.
Just about the only situation in which we view a pixel-per-pixel rendition of the original image is when we magnify it very large on our monitors. (And even this is more complicated now that we have monitors with very high pixel resolution.)
Additionally, when we compare images from different original pixel dimensions we often do apples-to-oranges comparisons without thinking about it. Let's say we look at two 1000x1000 pixel samples from two different sources. Let's say that one of them comes from an image that is 4000 pixels wide and the other from an image that is 8000 pixels wide. In the first case you are looking at a much larger portion of the image and magnifying it less. in the second case you are looking at about 1/4 of the area of the first image, and at a much greater magnification... so the effects of lens characteristics/ flaws loom much larger.
There's no idea way to get around this with digitally recorded images. However, one good approach is to look at real-world outputs that you might actually create with the images — rather than just clinical super-magnified computer screen images. For example, I think it is very useful to take the source images and go through the workflow to produce a best quality print at some size, let's say 30" x 45" with full frame. Actually make the print if you can... or take a letter-sized slice out of the two images and print that.
Alternatively, pick a common web display size that you might use and go through the workflow to produce the best possible outputs at that size... and then compare.
Using that first method — the 30" x 45" print — I think a lot of people might be surprised by just how hard it would be to distinguish the source of the image file if we compared, say, three: a 100MP miniMF sourced version, a 50MP or higher FF sourced version, and the 800MP pixel-shifted version....Show more →
|
