This post is spawned by what I did in this post:
http://forums.dpreview.com/forums/read.asp?forum=1010&message=31147076
Looking at High ISO 100% crops can be ugly. What we should keep in mind though is what a 100% crop represents.
I have measured the width of my monitor and using the known pixel dimensions I can calculate my monitor's dpi. That is pixels/measured_width or 1920 pixels / 20.4 inches. So my monitor dpi is 94.
That means when I look at a 100% crop from a G10 image I'm looking at an image that would require a 47 inch wide monitor to display the whole image (4416 pixel wide image / 94 dpi). I have never printed that big - I don't think most of us have.
So how close should we be looking? Well if I'm printing at 7.5x10, I should be looking at an image that is 10 inches wide on my monitor - in pixels that is 10 inches wide x 94 dpi = 940 pixels. I should either resize to 940 pixels wide or look at the full size image at about a 21% zoom level - not 100%. A zoom level of 25% would probably be close enough to give me a good idea of the printed noise.
Sometimes I print at 17 inches wide. In that case, I should resize to about 1600 pixels or zoom the full size image to 36%. Or 33% would be close enough.
It makes a big difference.
What follows are some 100% crops from a G10, LX3 and D3. And then a look at the same images in crop sizes that represent what I might print. I downloaded some ISO 800 raw files and ran them through ACR. The raw files are provided by Imaging-Resource.com and can be found here:
Imaging Resource G10 ISO 800 RAW File
Imaging Resource LX3 ISO 800 RAW File
Imaging Resource D3 ISO 800 RAW File
First the unprocessed (no NR or sharpening) raw files from the LX3 and G10. Both were opened with ACR using Adobe Standard profile, no ACR NR, no ACR sharpening. The LX3 image was upsized to match, using bicubic smoother. I also applied a -20 saturation in PS to all images.
Contrast those to this image from the D3 (one of the best high ISO cameras today):
From a darker area of the image:
And the D3, same area:
The P&S cameras look ugly in comparison. I think we all expect that. Also now that we've seen what the D3 can do, the P&S cameras don't look all that different from each other.
Next the same dark area crops processed. I used Noise Ninja, no sharpening, lum 7, chroma 9 (the crops were NN profiled, not canned profiles). I then sharpened with Nik 3.0 at 30% (Output).
The D3 processed:
If I had to grade them I would give the D3 an A and the other two a D. The P&S cameras do not hold up well at 100%.
Let's look at them at 10" inches wide (940 pixels). In order, the G10, the LX3 and the D3 (all processed):
If I had to grade these I would still give the D3 an A. The other two look like an A- or B+ to me. What do you think?
How about at 17 inches wide? I've resized the images to 1600 pixels wide and then taken 800 pixel crops from the middle. In order, the G10, the LX3 and the D3 (all processed):
Now I can see more noise in the G10 and LX3 but it's still not offensive (to my eyes). So I would grade them D3, A and the G10 and LX3, B- or C+.
Looking now at the 17 inch crops, the P&S images could have used a bunch more NR. I think the detail would have held up okay. I bet I could have gotten a solid B with a little more work.
So it was an interesting exercise to make these comparisons and I think I've made two points.
1. The G10 and the LX3 can make very good high ISO images and that will be obvious if you look at them with the appropriate zoom.
2. In the real world (at the sizes you're likely to print), the G10 and the LX3 are not that different at ISO 800.
I have printed the ISO 800 G10 and LX3 images at 7.5x10. They look very good and they are nearly impossible to tell apart.
I have read in several posts/articles that the G10 is a good "daylight" camera and the LX3 is a good low light camera. I strongly disagree with this assessment. If the G10 is only a valid choice before 7:00pm (because of the available light), then the LX3 is only a valid choice before 7:05pm. They are that close.
