vertical banding/gain variation at low ISO seems to be the issue with the 7D while horizontal banding/pattern noise at higher ISO is the issue with the 50D/5DII, so you wouldn't be using the deband on the 7D at high ISO anyway.
What is the best NR software for high iso 7D files?
Is Dfine better than Noise Ninja? Also, are you doing NR on the RAW file or on the TIFF after RAW conversion? Any difference?
alundeb wrote:
Lastly, I will include an example of ISO 100 midtone banding, 100% crop, here it is horizontal because this is a portrait oriented landscape shot. :D
we are just looking for any and every possible fault proven or not...why?
I dont understand how some of you come up with this stuff IE above picture
midtone banding? you got to be kidding me?also where do you learn such things anyway? I mean is there books,videos on pixel-peeping?
Technically, it's there. It does give the 7D about 2.3 dB less SNR at ISO 100 as measured by DxO, and those who base their buying decisions on DxO ratings instead of images, will think 2.3 dB means a lot.
so does this mean it cant take pictures?unless the all mighty DxO says so?
digitalbug30d wrote:
we are just looking for any and every possible fault proven or not...why?
I dont understand how some of you come up with this stuff IE above picture
midtone banding? you got to be kidding me?also where do you learn such things anyway? I mean is there books,videos on pixel-peeping?
Hello, we are not looking for faults, the background for this little trip was the surprisingly low SNR at low ISO the 7D got at DxO. We now explained exactly why, and how little it means, so that the 7D haters won't draw any false conclusions about small pixels being worse, etc.
Actually, I would expect critisism for defending the 7D, not the opposite.
Edit:
digitalbug, would you believe it, there are books on this, I studied digital image processing in 1993 and my work now occasionally includes hardware and software for infrared image capture and processing.
digitalbug30d wrote:
Technically, it's there. It does give the 7D about 2.3 dB less SNR at ISO 100 as measured by DxO, and those who base their buying decisions on DxO ratings instead of images, will think 2.3 dB means a lot.
It all depends on the scene and the information you are trying to extract from it ... 2.3 db can be a LOT to give up (especially if there is a systematic component to the noise) ... you never know. That said, even the 10D (which from the standpoint of noise isn't even in the same league as the cameras we are now discussing) is a great camera and capable of capturing terrific low ISO images the vast majority of the time. So what's a "vast majority" ... oh, about 99.9 percent of the time IMHO.
Except for the highly subjective "DxOMark Sensor" metric I really like what DxO is doing; however, IMO they would better serve visitors to the site by providing a some sort of visual perspective (on a per-picture basis) regarding what the differences in their numbers mean.
jkurkjia wrote:
Except for the highly subjective "DxOMark Sensor" metric I really like what DxO is doing; however, IMO they would better serve visitors to the site by providing a some sort of visual perspective (on a per-picture basis)...
Yes, like other informative camera review web sites do, and even helpful amateurs on this forum.
If DxO set out to show that a scientific approach to reviewing cameras would provide a more reliable benchmark, then it has failed at the first hurdle IMO. Other sites which post actual images, and especially those which ensure equivalent magnification do a better job of informing camera buyers.
I'm not saying that it isn't possible to devise a set of measurements and graphs which comprehensively show what needs to be known about comparative camera performance, I'm just saying that I don't think DxO has done so.
alundeb wrote:
To once again duplicate theSuede's findings, the vertical gain banding is static, what confused me about it was the black frame banding, that is not static to the same degree.
This means that it is possible to calibrate each copy of the 7D to remove all of the vertical banding in midtones/highlights. It is probably not going to happen in firmware, since the "processing" of raw files is done in hardware in the Digiq 4 processors, and additional processing in firmware would slow down the frame rate significantly.
(Edit: The paragraph below is false, the banding amounts to 9/2048, not 1/2048. I was too quick there. I will let it stand, to admit my error and apologize.)
For what I know, the column gains are already calibrated, but the precision of the calibration is limited to 10-11 bits (the banding amounts to about 1/2048 as of the signal, as standard deviation) . This is likely to happen if the internal processing precision is 24 bits. One would typically implement such a calibrated correction as multiplication by a parameter unique for each column, and division by a constant by means of shifting down a fixed amount of bits. To calibrate 14 bits data to a maximum error of 1/1024, one would then need 24 bits.
(Edit: The paragraph above is false, the banding amounts to 9/2048, not 1/2048. I was too quick there. I will let it stand, to admit my error and apologize.)
It is fully possible to preprocess RAW files with such a calibration. Hopefully the community can contribute to it, I don't know if it is "allowed" to tamper with CR2 files, or are they CRC protected?
Lastly, I will include an example of ISO 100 midtone banding, 100% crop, here it is horizontal because this is a portrait oriented landscape shot. ...Show more →
wow that is crazily mild!
you must have better examples than that!
i know i've seen far, far, far worse than that, which is almost impossible to see
brainiac:
....not saying that it isn't possible to devise a set of measurements and graphs which comprehensively show what needs to be known about comparative camera performance
I think that you are actually saying something similar. I believe that not only you but some others as well, believe their own tests or observations first and only agree if the "more scientific tests" support those findings.
It all reminds me of my old days when I was deeply involved in High End audio. Some would only believe their ears and nothing else.
mfurman wrote:
It all reminds my of my old days when I was deeply involved in High End audio. Some would only believe their ears and nothing else.
You are forgetting there's also this group of people who want to feel vindicated they have the made right choice after dumping a camera which is getting many people excited.
FWIW here is my DR chart comparing the 50D, 7D, 5D and 5D2 (i had to take the ISO normalizations from DxO data, sorry for sort of stealing that, visit Dxo, they are GREAT! www.dxomark.com ):
note this says nothing about whether ugly banding/pattern noise of one sort of another makes this irrelevant to an extent or not, the effective, non-objectional to the eye relative DR between may turn out to be differemt
note also these were normalized to 8MP as DxO does for their print comparison
oh and here is a photo quick shot off the deck, zero luminance NR and lots and lots of sharpening, keep in mind, ISO 400, 100% crop of a bird that was pretty far away (this with 300+1.4+7D and still this small ), 100% crop: http://skibum4.smugmug.com/photos/744174391_NWBqD-X3.jpg
thw2 wrote:
You are forgetting there's also this group of people who want to feel vindicated they have the made right choice after dumping a camera which is getting many people excited.
I could give a quack about how excited others are. It is my dollar & if
I am not happy with what I see my dollar stays in my pocket-
YMMV
, I don't see banding either.
