Originally posted by mikeodial Isn't this process different than just stacking, which increases image quality by software reinterpreting the image because it has better (more) image data. Whereas the PS technology is actually taking images that are positioned differently then merging them. This is more like interpolation, so the actual images being taken are from a different (albeit minute) new position, then merged in camera. I am pretty sure the result would be different than just stacking. The files should be smaller as well, (than 5-10 merged files for example), as they are still 24mp at the end of the day. However, I am not an image processing expert or camera designer, so could well be wrong. If it was a simple merge function there would be little value to this process don't you think?
In theory, we assume that pixel shifting in 1.5 pixels steps in x and y directions (assuming the sensor shift can accurately controlled within 1.95um in X and Y direction...which I highly doubt) would be equivalent to 2x oversampling. However, in practice the image falling onto the sensor is already band limited by the lens, so increasing the equivalent resolution of the sensor does not help as much as we may think (basically, what is oversampled is already having low contrast due to at least diffraction from the lens). According to simulations, pixel shifting increases resolution by 12% @f2, 8% @f4 and nearly 0% at f8 (that's why the DA*55 f1.4 lens is used to capture the pixel shift sample images on Ricoh web site). Now, lets assume that the lens out-resolve the 24Mpixels sensor. If what is advertised by Ricoh (the shifting is done in steps of 1 pixel) is true, this means that the spacial resolution is not increased. However, for sure the noise will be reduced, which give more room for sharpening, or more refined image. I'm eager to compare a 4 x16bits TIFF stacked and sharpened image from K-3, with a K-3II pixel shift photo in the same conditions (same lens, lighting etc..). This does not make the K-3II "pixel shift" unattractive, the advantage of the K-3 II is that you have the sharper image already boiled for you, hopefully available in raw. I look forward to do a comparison.