08-07-2011, 10:45 AM
[quote name='Tiz' timestamp='1312713184' post='10446']
"The edges are what needs "correcting" by stretching them, basically. This leaves the center more or less unaffected. You end up with a larger image, pixel resolution wise. Then you crop this to original pixel resolution."
This is not what's happening with the Canon SX220. I have not tried any other cameras. The maximum resolution of the sensor is 12 MP (4000*3000). This is the distorted image. If you now start to correct the distortion, you will have to crop a certain amount of the edges and, consequently, the cropped image has less than 4000*3000 pixel (can be easily tested if you work with RAW files). If you shoot .jpg the camera nevertheless saves a corrected 4000*3000 pixel image. But this can only be because the image was extrapolated (aka bloated back to the orginal size). It might be a bit different with a 12MP sensor that can actually record more than 4000*3000 pixel.
[/quote]
You do not get the principle yet.
With the distortion, the corners/edges are "squashed". Not the center.
So... if you stretch the edges, to correct the distortion, you end up with a BIGGER image, pixel resolution wise. Which you then can crop to original pixel resolution. The center "optical" resolution remains mostly untouched.
You can NOT test that easily with RAW files, if you only have software that corrects if by shrinking the center of the image, instead of expanding the edges. So, not sure why you think you know what your SX220 does internally. All you know is what the RAW converter does that you are using.
With expanding the edges, you lose sharpness on the edges. With shrinking the center you lose sharpness in the center due to the shrinking, and then later too if you expand the resulting image to the original pixel resolution. Expanding the edges/corners is the smartest thing to do, and that is most probably what Olympus and Panasonic are doing.
The MTF measurements to suggest that too, with the edges trailing the center resolution with a considerable and quite uniform over the aperture range difference.
"The edges are what needs "correcting" by stretching them, basically. This leaves the center more or less unaffected. You end up with a larger image, pixel resolution wise. Then you crop this to original pixel resolution."
This is not what's happening with the Canon SX220. I have not tried any other cameras. The maximum resolution of the sensor is 12 MP (4000*3000). This is the distorted image. If you now start to correct the distortion, you will have to crop a certain amount of the edges and, consequently, the cropped image has less than 4000*3000 pixel (can be easily tested if you work with RAW files). If you shoot .jpg the camera nevertheless saves a corrected 4000*3000 pixel image. But this can only be because the image was extrapolated (aka bloated back to the orginal size). It might be a bit different with a 12MP sensor that can actually record more than 4000*3000 pixel.
[/quote]
You do not get the principle yet.
With the distortion, the corners/edges are "squashed". Not the center.
So... if you stretch the edges, to correct the distortion, you end up with a BIGGER image, pixel resolution wise. Which you then can crop to original pixel resolution. The center "optical" resolution remains mostly untouched.
You can NOT test that easily with RAW files, if you only have software that corrects if by shrinking the center of the image, instead of expanding the edges. So, not sure why you think you know what your SX220 does internally. All you know is what the RAW converter does that you are using.
With expanding the edges, you lose sharpness on the edges. With shrinking the center you lose sharpness in the center due to the shrinking, and then later too if you expand the resulting image to the original pixel resolution. Expanding the edges/corners is the smartest thing to do, and that is most probably what Olympus and Panasonic are doing.
The MTF measurements to suggest that too, with the edges trailing the center resolution with a considerable and quite uniform over the aperture range difference.