Sunday, May 9, 2010

F70EXR – Larger versus Medium Size – Yet Another Test

The EXR sensors have afforded us a new way of shooting. Hardware assisted dynamic range (DR) extension provides far better captures in the day time than can be obtained by lesser sensors. This because they can split the sensor into two separate sensors and stop the exposure early for the highlights. The combined image can store a much wider dynamic range than any other small sensor.

This binning mode (combination of two adjacent like-colored pixels in the case of the EXR technology) also has the advantage of reducing noise as ISO rises.

But the EXR technology can be ignored in a high resolution mode that uses all of the pixels as individual picture elements. I have railed against this use of the cameras on any EXR model because I have looked at images produced in both modes from all of the EXR sensors (F200EXR’s larger sensor, shared with the S200EXR, F70EXR’s smaller sensor, and F80EXR’s smaller and denser sensor) and none of them retain more low contrast details in hi res mode. In very bright light, they can retain slightly more detail with very bright details, but edge integrity still suffers as the noise reduction that must be used on hi res images kicks in.

I’ve shown this several times and you can search this blog for the articles if you would like to see the tests.

But … a fellow named Oasis Feng left a comment on my how to shoot article for EXR sensors that he had discovered that the 10mp L size mode for the F70EXR is not completely useless. He shot a long exposure with a tripod in a dark environment and he sees more fine details with the large image.

This, of course, forces me to test that specific example, since none of my L vs M tests have done exactly that.

The short answer is that M upsized to L size still defeats the native L size. It’s quite subtle, in that the L size shot at 100 iso (when using tripod, there is little need to raise ISO unless 8 seconds is insufficient for your exposure) retains slightly better noise character. This because I double the size of the noise particle on upsizing the M image. But the advantage there is pretty small.

I shot a tooth brush in a dark bathroom at 8 seconds, and that resulted in a slight underexposure that required lifting in photoshop. I cropped the two out of camera images (after upsizing the M image to match) and put them together before performing any other processing. I lifted the exposure, added a bit of contrast, and sharpened the crops together.

The arrows in the following image show where I prefer one over the other for detail retention. The L image does have a couple of better spots. But the M image has more.

It’s quite close, so don’t get too excited about it. It is obviously not going to make much difference at web sizes, but as with any small sensor, low contrast details such as those in shadow, are lost earliest. If you shoot in M mode, you will always hold them longer. There is no way around that. Lower density wins.

Click through to see the full sized crops.


Scotty said it best: “Ye canna change the laws of physics!


Oasis Feng said...

Thank you very much for such a quick but sufficient test as response.

According to your conclusion, I reviewed my test photos once more, and found that the better details were actually only in brighter area. As I turned to the darker area, I did find more subtle worse details this time, exactly as you mentioned.

Sorry for the partial judgment in my earlier comment. I'll do my tests more carefully and thoroughly next time. :)

Kim Letkeman said...

No need to apologize at all. In fact, the differences are very subtle and would not likely show up on a web sized image unless there were a lot of them in the same area. Thank *you* for alerting me to this test case, as I would not have thought to test it had you not mentioned it.