How do I delete multiple rows and columns in an image?

Question

I am looking into how digital SLRs extract video data from a CMOS sensor. To give an example, the GH1 from Panasonic can record video at 1920 x 1080 from a 12 MPixel sensor that, say, is 4000 horizontal x 3000 vertical. How they do that seems to be not in the public domain, but there have been a few guesses (see http://www.dvxuser.com/V6/showthread.php?t=206797 and http://luminous-landscape.com/forum/index.php?showtopic=38713).
One approach would be to simply read every second row of sensor pixels (1500 rows read from the original 3000) and once you have those in memory, delete every second column (2000 columns left). You would end up with a 2000 x 1500 image which could then be resampled to 1920 x 1080.
I'd like to simulate what the camera appears to be doing, by generating a 4000 x 3000 test image and then asking Photoshop CS4 to delete the appropriate rows and columns. It may not necessarily be every second row; the Canon 5DMk11 appears to read every third row, so I may need to delete two out of every three rows.
Can Photoshop do that sort of thing? If so, how?

Answer 1

Thanks for the suggestions. Yes, I did take a detailed look at your images, but they weren't 100% convincing because it wasn't clear just what was happening. And Adobe's explanation, after reading it again, explains nothing at all to someone who doesn't know how Nearest Neighbor works.
But you are correct -- Nearest Neighbor does effectively delete pixels. I proved it with the attached 6 x 6 image of coloured pixels (the tiny midget image right after this full stop -- you'll have to go to maximum zoom in PS to see it).
These are the steps to delete every second row and then every second column.
1. Select Image > Image Size.
2. Set Pixel Dimensions > Height to half the original (in this case set to 3 pixels).
3. Set Resample Image to Nearest Neighbor.
4. Click OK and the image shoould now consist of three vertical white strips and three vertical Green-Red-Blue stripes.
5. Repeat steps 1-4, but this time set Pixel Dimensions > Width to half the original (in this case set to 3 pixels). The image should now consist of three horizontal stripes Green-Red-Blue.
Just to make sure the method worked for every third pixel, I repeated the above steps but with 2 pixels instead of 3 and obtained the correct White-Green, White-Green pattern.
I resampled the Height and Width separately for the test so that I could see what was happening. In a real example, the height and width can be changed at the same time in the one step, achieving the same results as above.
Finally, how to explain how Nearest Neighbor works in the simple cases described above?
Division by 2
In the case of an exact division by two (pixels numbered from top leftmost pixel), only the even numbered rows and columns remain. To put it a different way, every odd numbered row and column are deleted.
Division by 3
Only rows and columns 2, 5, 8, 11... remain.
Division by N
Only rows and columns 2, 2+N, 2+2N, 2+3N... remain.
To put it simply, a resample using Nearest Neighbor (using an exact integer multiple) keeps every Nth row and column, starting from number two. The rest are deleted.