I posted about this week or so ago
here (post #8). Pentax has all the technological parts already since they "dither" the sensors position, and in their in camera HDR they combine the images. All they need to do is to apply them in a slightly different approach. What is different here is that you know beforehand the direction and the amount of movement (left, up, right right, down down, left left, with each set of movements of about 1.5 pixels),. What you want to do is to fill in the gaps within the Bayer pattern to increase the resolution, and or shift the Bayer pattern over a different pixel color, so as to mimic the Foveon sensor. I would think that with 4 images you could fill in the holes and with 6 images, you could achieve a stacking of the pixels.
Actually, the implementation could be done to either fill in the holes, or stack the Bayer pattern with different colors shifted into the same position - i.e. the Foveon sensor. The user could possibly have their choice - i.e., two additional modes.
It would definitely take longer since you would have 6 discrete images to capture (the original with an additional 5 dithered in various directions). With SR the time to dither the sensor would be nearly instantaneous. The one aspect that I would hope that they would change is how they combine the raw images into the resulting image. Two approaches come to mind.
- In camera - combine the raw images into a single resulting image (hopefully in a RAW format and not just a JPG like in the current HDR implementation. Since this is a composite of 6 individual images into a single resulting image (much larger), the current K5/II/IIs would probably not have sufficient buffering available.
- Out of camera - Have the camera record the individual raw frames similar to automatic Bracketing (with additional annotations in the EXIF, direction and amount of the dither), then within post processing have a utility combine the images and produce the resulting image (I would hope as a TIFF or DNG). This approach would certainly solve the buffering size problem, but would probably need a customized post processing utility provided with the camera.
I would certainly think that this could be an additional shooting mode. It would certainly - I think be popular with landscape photographers. This is just a slight variation in the way you stack images.
I would believe that this is very similar to, but an ever so slightly different approach to what is used in image exploitation. A combination of various sensor types, for example visible light imagery with IR and or SAR (Synthetic Aperture Radar) imaging. The difference is shape and size of the "pixels", in that visible light pixels are usually square and small, while IR pixels are rectangular and a bit larger, so the alignment tends to skew the results. SAR which is a single composite line of pixels, that are striped together and run as a "roll of toilet paper" for an indefinite length.
Astronomers stack images from different sensors all the time.