How are "smaller" than 16MPx pictures obtained (such as 2MPx video)?

Started Oct 8, 2011 | Discussions thread
 Forum Parent First Previous Next Next unread
 Like?
 How are "smaller" than 16MPx pictures obtained (such as 2MPx video)? Oct 8, 2011

Most cameras offer 2 or 3 different "size" recording options. Large, is either the full 3:2 or 4:3 sensor PIXEL COUNT. There will be a similare set of pixel counts for a 16:6 aspect ratio on the sensor.

Here's my puzzle. I assume that, except for the 16:9 window, that the smaller images can not be smaller windows on the sensor because if they were, then the smaller pictures would have different fields of view (and DOF).

The OBVIOUS way smaller pixel counts could work is:

1) down-scale each row via interpolation.

2) skip rows and columns of pixels. to obtain fewer pixels from the sensor.

The problems I see with these solutions are:

1) Because there is a Bayer pattern on the sensor, skipping would need to be done in a way that the reduced pixel count frame still has an in-tact Bayer pattern. I think this forces reduction by a factor of 2 -- at a minimum. But, the smaller size pictures are not that much smaller. So it seems, besides the aliasing problem, that's not how it's done. Correct?

2) To downscale columns, a DSP would need to have a full buffer from which it would interpolate both horizontally and vertically into a second smaller buffer. This would be done AFTER debayering so there the big buffer would be RGB. (Obviously, not 4:4:4 color sampled RGB because the sensor can't deliver 4:4:4.) Is this how it's done?

===========

When recording video, only 2MPx are needed -- which is far fewer than even the smallest photo pixel count. (On the NEX-5n, the smallest photo is about 5MPx.) Does anyone know if the YCrCb data is pulled from the SMALLEST photo RGB buffer, or does video take a different path from the sensor?

For example, a 16:9 window would be EITHER debayered and downscaled to 2MPx OR columns and rows would be skipped (while still keeping a Bayer pattern) to create a smaller frame that is debayered and further scaled to 2MPx. The obvious advantage of the latter approach is that the data read-out from the sensor -- at 60p -- is one quarter that used in a photo.

This would also explain why stills captured while shooting video are -- on some cameras -- 1920x1080.

Complain
 Forum Parent First Previous Next Next unread
Post ()
 Forum Parent First Previous Next Next unread
Keyboard shortcuts: