jorkata wrote:
Hi UCSB,
This is very encouraging.
Do you think you can post links to some of the RAW files from this comparison.
Maybe the ISO-800 files from both cameras.
Thanks.
I would be happy to post the files. But, I have a little problem. I host my files on my Comcast account and they have a 8 MB limit. I'm fighting the flu today and don't have the energy to go out and try and find a hosting solution. When I am feeling a little better, I come back and post them.
wayne seltzer wrote:
Hi UCSB,
Which firmware version did you have when you had the problem and which one do you have now?
I just got back from a trip to southeast asia and now got to play with my new 7D which I got just before the trip but did not bring as I needed the wideangle abilities of my FF 1ds3.
After only shooting a couple dozen pictures I am finding much better color response in the greens compared to all my old cameras(1ds3,5D,1D2N). I also am finding the automatic WB to be very good indoors with fluorescent and incandescent lighting, something my old cameras had problems with.Not a big problem since I shoot RAW all the time but will save time in adjusting it in post all the time. Also, the exposure is much more accurate because its color aware.I am also seeing less of the washed out color look with this camera compared to my old ones. The colors look more vivid to me. This is just looking at the replay of the pictures on the camera screen and I haven't yet processed the RAW files with DPP yet,
so I am not sure yet if what I am seeing so far is due to better files or just a better display screen on the camera.
AF tracking so far was very responsive like my 1d2n and very accurate on a couple of flying bird shots.
I need to test it out alot more extensively.
I haven't seen any banding yet but will be on the lookout for it.
The only negative I see so far is more high iso noise compared to my FF 1ds3 but that is to be expected....Show more →
I have v1.0.7, I now have v1.1.0. Have fun with the new camera.
ejmartin wrote:
UCSB, glad to hear it's working for you now. Wonder if one of the firmware upgrades implemented a silent fix.
interesting
i'm a little surprised they would make that silent though since they never made any of the black dots/banding in mRAW fixes for the 5D2 silent (IMO those issues showed up far less than the mazing too, i never even saw the mRAW banding once myself)
Anyway, if a firmware fixed it, it is definitely not 1.09 since my new body came with that version installed. I haven't tried 1.10 yet. I'll give that a go tomorrow. Hopefully it will work out. Too bad I didn't load it before the World Series.
ejmartin wrote:
UCSB, glad to hear it's working for you now. Wonder if one of the firmware upgrades implemented a silent fix.
nope
Firmware 1.07, 1.09. 1.10 (all that have been released) all act the same way with ACR 5.5-style demosaicing, here is a 1.10 crop (ISO 100 of a textureLESS and out of focus red square):
I wonder if UCSB's mazing really went away, that sounds pretty bizarre to just disappear like that, maybe he just shot the wrong subject with the last shoot that by chance featured only things least likely to make it really obvious
Here is the same ISO 400 test shot converted in ARC (top) and C1 (bottom). Shown at 100%. But, I went in and zoomed to higher magnifications and you are getting a very small amount of patterning (same in C1). Most people would consider the patterning ISO 400 noise ... nothing like what you are showing above in pink. I think everyone knows that ARC is not working so using it for testing does not make sense. I don't really understand why my prior problems disappeared. Maybe take all of the batteries out of your camera and reset it, or call Canon and see if the camera has a reset function.
UCSB wrote:
Here is the same ISO 400 test shot converted in ARC (top) and C1 (bottom). Shown at 100%. But, I went in and zoomed to higher magnifications and you are getting a very small amount of patterning (same in C1). Most people would consider the patterning ISO 400 noise ... nothing like what you are showing above in pink. I think everyone knows that ARC is not working so using it for testing does not make sense. I don't really understand why my prior problems disappeared. Maybe take all of the batteries out of your camera and reset it, or call Canon and see if the camera has a reset function.
I will shoot my Color Checker SG next and post that. It contains a very wide range of colors.
...Show more →
Be sure to follow the procedure that skibum does: slightly out of focus, push sharpening way up, and view at 200%. If there is mazing, you'll see it this way.
Fred Tedsen wrote:
Be sure to follow the procedure that skibum does: slightly out of focus, push sharpening way up, and view at 200%. If there is mazing, you'll see it this way.
Yeah, just like we take, process and view real photos.
UCSB wrote:
I would be happy to post the files. But, I have a little problem. I host my files on my Comcast account and they have a 8 MB limit. I'm fighting the flu today and don't have the energy to go out and try and find a hosting solution. When I am feeling a little better, I come back and post them.
You can upload them to a hosting service like yousendit.com. It's free for both sender and receiver. The RAW of a slightly OOF colorchecker would be useful.
keithreeder wrote:
Yeah, just like we take, process and view real photos.
I despair...
I've seen mazing in "real photos" from prior x0D cameras that I have had to push a bit in post-processing. I didn't go looking for it... it found me. I'd rather the mazing not be there.
I just got my 7D. I appreciate these discussions, so I can look to see if this is an issue for my camera. Hopefully it isn't. If I have an atypically bad camera, I'll exchange it. If they're all like that, I'll keep it. If it's an inherent "feature" of the camera, the discussions may help folks writing raw conversion algorithms (e.g., ejmartin) do a better job. They may help us better understand how to deal with the image in postprocessing of "real photos" where this is an issue, even if such photos are rare.
Do you "despair" because you feel people are wasting their time? Your time? Bandwidth? Because you think they are besmirching the reputation of a fine camera, and people are holding off their purchases when they shouldn't?
As for the last potential reason I listed: it would be one thing if we were writing an official review for a huge website, and recklessly scaring naive first-time buyers about the latest xx0D... but that's not the case here. Anyone considering plunking down their US$1700 or whatever GBP probably should consider the purchase carefully, and evaluate all the information on balance. I don't see anti-7D hysteria here.
gfiksel wrote:
Well, here is a Color Checker shot at ISO 100, defocused.
Developed in LR3beta, at 100% and 200%, sharpness at 25 (default) and 150 (maximum)
In addition, mRaw is shown for comparison.
What can I say... Yes, at 200% and sharpness all way up one can see the mazing. At normal sharpness level and magnification it appears to be normal. The mRaw option exhibits it even to a lesser extend
Your jpeg compression level of 6 has inadvertently done a bit of NR for you; that is why the 100% view doesn't show the artifacts as much as 200% view, and even the 200% view has smeared them substantially (BTW this is why I use png's for this sort of demonstration -- there is no compression in png's, it is a lossless format; I wish this site would enable its software to display png, since it is a common, open format for web display).
mRAW will iron out the artifacts because it is downsampling below the scale at which they are resolved. It has always been the case that the artifacts can be avoided by smearing the data and lowering the resolution. For me the issue has always been, is there a way to avoid the artifacts while maintaining the full 18MP of resolution.
abqnmusa wrote:
so it would appear that over-sharpening causes mazing
NO. That is only in ARC, I am not seeing that in C1. We know that ARC is not working correctly and the 7D is not fully supported. It is a waste of time to use defective software to perform tests.
Today, I shot my Color Checker SG at 100, 400, 1600, 3200 (both focused and defocused). The Color Checker SG has all of the colors of the regular Color Checker and many, many more. I did my processing in Capture One PRO v5. I did not see any mazing. There is noise present (even at ISO 100). If you crank up sharpening, it becomes very much more pronounced. I am not seeing anything different than the "snowman" images/crops I've already posted.
Fred Tedsen wrote:
Be sure to follow the procedure that skibum does: slightly out of focus, push sharpening way up, and view at 200%. If there is mazing, you'll see it this way.
although to be honest with that red patch above, in that one, I simply used the same sharpening settings throughout that I do for actual photos I'd take with my old 20D, the sharpening is getting towards the high side but it's not at insane levels or anything
anyway as for why UCSB's suddenly got better, did he maybe use masking and NR in ACR? change the ACR pre-sharpening settings? not sharpen at all in CS4?
Indeed, now the sharpness 150, 100% crop of the full RAW shows the artifacting much more clearly; it was smeared before. The jpeg compression will smear lower contrast detail more.
If the jpeg compression weren't getting in the way, one should be able to sharpen the Sharp 25 crops and recover the artifacting of the Sharp 150 crops, but that ain't gonna happen since the jpeg compression has flattened the fine low contrast "detail".
This BTW illustrates one reason why 400% crops can be useful -- if people are posting jpeg images, the larger scale can get around the jpeg compression and reveal what's going on at the pixel level. But as I mentioned above, all this would be moot if a lossless format such as PNG were used.
keithreeder wrote:
Yeah, just like we take, process and view real photos.
I despair...
1. it doesn't even need to be out of focus to cause damage
2. many photos have smooth or OOF areas in them anyway
3. for that red patch I used a regular, if not small, amount of sharpening, the same amount I have used on many actual photos before with no issues of the like at all
4. yeah it is at 200% but that just makes it clear that it is not jsut an awful lot of noise in general, but mazing noise, for those with high dot pitch monitors
anyway, hopefully ACR 5.6 will do better (acr 5.5 stinks with it, DPP is not really all that great either, C1v5 is getting somewhat better)
abqnmusa wrote:
so it would appear that over-sharpening causes mazing
it makes it more obvious, but i have had it show up very noticeable with the same old sharpening, even my slighly less agressive mix, that i use for all my photos
heck how do you think i noticed it in the first place?
by processign a NORMAL photo using my NORMAL settings and going like oh man....
