Which is better: low resolution mode (sRAW) or downsizing afterwards?

[McQ]

Thanks for your examples. Very helpful.

edit: here is someone else's test - http://jasonography.blogspot.com/200...-test-40d.html

In this link the test images of the flowers seem to have been taken in error. There's a clear difference in the depth of field or position of the focal plane, which was not due to sRAW. Just my opinion.

[Colin Southern]

I'm quite interested in doing a little test on the Canon 1Ds3 to see if the noise from a single sRAW capture is equal to the noise of 4 averaged RGB captures.

How does this look for a methodology? ...

- Take 1x sRAW at highest ISO

- Take 4x RAW at same ISO

- Average 4x RAW shots & crop to appropriate sample size

- Upsample sRAW by 200% using nearest neighbor, and crop to same sample size.

If sRGB mode is using pixel binning then the noise should be very similar. Sound OK? (on the 1Ds3 sRGB exactly halves both pixel dimensions, thus providing a 4:1 resolution ratio).

[McQ]

I'm quite interested in doing a little test on the Canon 1Ds3 to see if the noise from a single sRAW capture is equal to the noise of 4 averaged RGB captures.

Are you wanting to compare noise at 100% or at a given print size? If it's the former, then I would just do an equivalent XbyY pixel crop for both the sRAW and averaged RAW without any up/downsizing. If it's the latter, I would downsample the averaged RAW instead of upsampling the sRAW.

If the goal is to have nearest neighbor mimic/verify the pixel binning look, the problem with this is that pixel binning is performed prior to the RAW conversion, but yet the images are being shown after RAW conversion. I'd expect them to look different.

[kevinf]

I would downsize the RAW 50% with bicubic. Nearest neighbor will trash smooth toned images.

[Colin Southern]

Hi Sean,

I'm trying to think of a way that compares apples with apples; if I don't adjust the resolution somewhere along the way then (at 100% magnification) the sRAW image will appear to be zoomed in twice as much, and yet if I down-sample the RAW shots then that reduces the noise anyway, independently of any averaging ... which only left the up-sampling of the sRAW - not sure what effect that has on noise though.

... definately one of my weak areas!

[McQ]

Yes, the magification difference is definitely a valid point. However, if you take out of focus pictures of a reasonably textureless solid color or grayscale subject that shouldn't matter. My favorites are a wall or a sheet of paper. Pretty boring though, I know

[kevinf]

From what I'm seeing out of 40D sRAW, it gives a substantial improvement with ISO 3200 regarding noise. Otherwise, the noise levels are quite comparable in actual images below ISO 3200. What is interesting is the much sharper image sRAW gives. Quite a substantial difference, upsizing the sRAW to RAW size still gives a sharper image, albeit with less total detail.

So if I were going for a tack sharp image, I would use sRAW.
If I was shooting at ISO 3200, I would use sRAW.
If I was shooting to create images to be viewed on a monitor or HDTV, sRAW.

If I would be printing poster sized, RAW.
Shooting below ISO 3200 for as yet undetermined use, RAW.