I guess I have to be the first to say it... at the pixel level those ISO 3200 samples (without adjustmets) don't seem better than the current 5D, and may be even worse. Neither do they look better than the 1Ds3, as Canon marketing literature has suggested.
This leads me to wonder whether for this camera, as with the 50D, Canon's claimed noise improvement over the predecessor model is due to the use of stronger NR settings in-camera and in DPP?
Has anyone else noticed that in DPP 3.51 if you leave the default NR preference set for "use camera default" there is some NR applied to *all* images, even when "High ISO Noise Reduction" is disabled in the camera? I think the idea behind these defaults is to equalize DPP processing with the in-camera JPEG processing, which always applies a little NR. I tend to override this, as DPPs luminance channel NR is pretty destructive to fine detail in the image even at its lowest setting.
(And yes before Brainiac jumps in, I'm aware of that downsampling the higher resolution images will improve apparent noise performance when compared to lower resolution sensors... )
I've also found that, contrary to my previous understanding of chroma noise reduction's effects, DPP's chroma noise reduction is also destructive to fine detail, at least for G10 raw images. At iso 80 for the G10, setting chroma noise to 0 (vs. default settings) and, of course, luminance as well at "0," results in much sharper images, without any detectable chroma noise.
Having both a 5D and a 1DIII I can say that Canon has made a lot of improvements in imaging since 2005. The 5DII will have better noise characteristics and better overall imaging characteristics than the 5D. If you have a 5D you will see this immediately. Web posted shots may, at times, suggested that there is some question as to real improvements in the 5DII.
My definition of "meaningfully better" is "I can tell the difference in print". I am sure if one had mixed prints of identical scenes shot with the 5D and 5DII, the 5DII prints would easily be selected as better by everyone.
httivals wrote:
I've also found that, contrary to my previous understanding of chroma noise reduction's effects, DPP's chroma noise reduction is also destructive to fine detail, at least for G10 raw images. At iso 80 for the G10, setting chroma noise to 0 (vs. default settings) and, of course, luminance as well at "0," results in much sharper images, without any detectable chroma noise.
I don't think we can generalize G10 to DSLR. Although it sure looks like the G10 is a great camera.
dcmiller wrote:
Having both a 5D and a 1DIII I can say that Canon has made a lot of improvements in imaging since 2005.
Perhaps... though my experience with both the 40D and 50D suggests that Canon has gone backwards somewhat in terms of both noise and per-pixel somewhat compared to the predecessor models in that line.
Incidentally, I prefer 5D files to those of the 1DIII and consider the 5D to be a high-point in terms of IQ for any make (2005 technology or not).
dcmiller wrote:
Web posted shots may, at times, suggested that there is some question as to real improvements in the 5DII.
Yes, true. In this case, though, the OP has posted unretouched full resolution images from DPP with no NR applied... it's the first time we've seen unmodified samples that get close to what the sensor can do without post-processing.
) 