atodzia wrote:
Thanks Hatch. One thing I have noticed is that after looking at a few shots I have taken full size in Photoshop, since receiving the camera yesterday, it is a little of a let down after they are downsized to web posting jpeg size and don't have quite the same impact At least on my 27 inch NEC monitor. I think subject matter has something to do with it, as well as jpeg downsizing methods.
You will not see any resolution advantage in web-sized photos from a super-high-resolution camera.
A few years back I asked readers to participate in a little experiment at my blog. I shared several versions of a photograph of a particular subject and asked them to compare. I did not tell them what the difference was, but I asked them to judge whether the images were the same or different.
Viewers were shown three images that had been downsized for screen presentation on the web. It was what might be called an "ABX discrimination test." Samples A and B were different and sample X matched on of the first two. Viewers were asked to determine all three were the same, all three were unique, or two were the same and one was different. In the end it turned out that the test indicated that viewers could not tell the difference between the samples in any consistent way.
100% magnification crops from the test images are available here: http://www.gdanmitchell.com/2009/12/06/experiment-1-the-answer
(You can see that the resolution of the sources images is grossly different — yet once they were resized for the web the differences were not apparent.)
This reinforced what seems obvious, namely that web display doesn't challenge the resolution capabilities of much of any camera.
gdanmitchell wrote:
You will not see any resolution advantage in web-sized photos from a super-high-resolution camera.
A few years back I asked readers to participate in a little experiment at my blog. I shared several versions of a photograph of a particular subject and asked them to compare. I did not tell them what the difference was, but I asked them to judge whether the images were the same or different.
Viewers were shown three images that had been downsized for screen presentation on the web. It was what might be called an "ABX discrimination test." Samples A and B were different and sample X matched on of the first two. Viewers were asked to determine all three were the same, all three were unique, or two were the same and one was different. In the end it turned out that the test indicated that viewers could not tell the difference between the samples in any consistent way.
100% magnification crops from the test images are available here: http://www.gdanmitchell.com/2009/12/06/experiment-1-the-answer
(You can see that the resolution of the sources images is grossly different — yet once they were resized for the web the differences were not apparent.)
This reinforced what seems obvious, namely that web display doesn't challenge the resolution capabilities of much of any camera.
I wonder if the downsizing method will be critical? I currently use bicubic smoother in one step followed by a sharpening using Topaz. I have used dozens of methods including some of the multi step actions. Years ago I settled on the simple one I use now.
But we will be going from over 8,000 to 850 wide, thats a big step.
Folks say "you can't learn anything from images posted on the web". But they are only right when the image has been resized. That's why I always try to include a "100% crop" of at least part of an image if the post is about anything technical or related to IQ/resolution/noise/sharpness etc. When you see an image in which the pixels are not original (i.e. the image has been resized) you simply can't learn about the properties of the original image - the ones related to the original capture and the capability of the camera/lens combo.
However, with a chunk of the image that has not been resized (aka a "100% crop"), you can directly compare the parameter of interest between images and learn a lot about the (comparative) performance of specific lens combos, side by side on a single display. That, of course, is not to say that's how we will view normally the image. Also, of course, we will take into account the key parameters of how big the whole image would be when viewed at 100% and how much the optical image has been enlarged from the sensor size when assessing both lens and sensor performance. Once you have done that, you can resize for comparative purposes, or make test prints etc etc to cunstomize your evaluation for your workflow, style, intended audience etc. to get back to "reality"
the 100% crops should look similar to what 7d2 produces. It is impressive seeing how much one can zoom in to get a 100% crop, and extrapolate to how much detai there is in a large print.
sometimes, improvements in features such as fps, AF and DR are easier to demonstrate thank an increase in resolution but its definitely there. Roger cicala showed a 20-30% increase in resolving power vs 5d3 in the center of image with the best of lenses (300 28L II, and zeiss 85 1.4), with a bit less in the corners
stanj wrote:
Well the last one is 400, as it says. The EXIF posted is accurate, except of course the camera model is not ">9"
I meant the last one in the post I quoted, the one that says ISO 100. Checked the EXIF and it says 100. Just thought the noise was a bit on the high side, very similar to ISO 400
ben egbert wrote:
I wonder if the downsizing method will be critical? I currently use bicubic smoother in one step followed by a sharpening using Topaz. I have used dozens of methods including some of the multi step actions. Years ago I settled on the simple one I use now.
But we will be going from over 8,000 to 850 wide, thats a big step.
The downsizing method has a far, far, far bigger effect on the final sharpness of web images than the original resolution has, at least when we are talking about files from cameras released in the last half decade or more. Sharpening methods are also important.
The detail produced by a 50MP file is gone in web presentation though. For example — using rough numbers — each pixel of a 1024 pixel width image on the web consists of the average of around 60+ pixels from the original 50.6MP 5Ds file.
alundeb wrote:
I meant the last one in the post I quoted, the one that says ISO 100. Checked the EXIF and it says 100. Just thought the noise was a bit on the high side, very similar to ISO 400
Ah yes. It's not SOOC, I had to boost it a bit, the original shot was quite underexposed (shooting into the setting sun, light reflecting on the water surface). Don't have the file here to see how much I pushed it.
molson wrote:
Yes... although it makes a big difference whether or not you have a lot of fine detail in the image.So processing shouldn't be too bad. I figured the files would be a bit larger. I was looking at some of my D800/E they were 70-78mb and weren't too bad to process. I guess just depends on comp specs
Dpedraza wrote:
So processing shouldn't be too bad. I figured the files would be a bit larger. I was looking at some of my D800/E they were 70-78mb and weren't too bad to process. I guess just depends on comp specs
Processing the 5D SR files is roughly 10x faster than Sony A7R files, despite the Sony files being so much smaller. There must be a lot going on in the background with LR handling all the baked-in settings in the Sony files. I haven't compared it with Nikon D800E or D810 files yet.
I had to keep checking to make sure I hadn't set my camera to JPEG by mistake...
ben egbert wrote:
Hard to see the improvements over 5D3 until you print large enough to use all those pixels.
... or crop enough.
One of the advantages I am realizing about the 5DS is that you can compose with extra room and then crop later. So, when in doubt about composition, pick the wider option and tighten the crop in post-processing.
At least on my 27 inch NEC monitor. I think subject matter has something to do with it, as well as jpeg downsizing methods.
Now I want one 
