More terrible sky noise with A700

[bakubo]

And have you compared RAW to cRAW?

No, I haven't. I have read several times though that the only difference anyone has ever been able to detect is when there are sudden, sharp transitions from very dark to very light such as specular highlights, etc.

[bakubo]

I've noticed the same sort of problem with the A350, in the same sort of shots, when the sky goes from darker to lighter across a frame, you get these distinct "borders" between the different shades.

which is different to the sky shot WestCoast posted there is not has large change is sky color across the frame.

Yes, I think you are right.

[bakubo]

Hi Henry. Does this happen for all blue skies, or is this a similar problem to the one DK pointed out the in A550 thread (see http://www.photoclubalpha.com/forum/vie ... =32&t=3331) when it is only sometimes ?

I don't know if it is the same thing or not. See my old thread I linked to in the OP of this thread to see where I raised this problem a few months ago. I don't think it happens with every blue sky, but it does, as stevecim wrote, when there is a transition of blue in the sky -- which happens a lot.

[RubberDials]

This looks like a gamut issue to me. There aren't enough colours in the colourspace to implement a stepless blend in the range of tones depicted so the processor is dithering.

[agorabasta]

Henry,

Your example clearly shows red clipping down in the sky. It's then further exaggerated by 'clarity' setting.

Here's what always works in the 'skies' for me - set recovery much higher at 70 or even more; set blacks to zero; if reds still clip down, add exposure. By all means avoid primaries clipping down in bright saturated areas. After that adjust the tone response by the secondary curve adjustments and brightness/contrast. Only then go to clarity/vibrance/saturation.

It should always work for you.

[agorabasta]

By the way, do you think that it is ACR or the A700 that has so much trouble getting the hue of blue skies correct? Your hue/saturation adjustment improved things a lot, but I am wondering why it is so off-kilter to begin with? Anyway, this is a side issue to my main problem at the moment.

The ACR always worked best for the Canon. The Canon red response has a wider spectral sensitivity curve, and the red curve also has a long 'shelf' protruding far into the green. This way they avoid red channel noise and down-clipping. But frankly, I cannot stand the 'Canon red' (and the sky-blue also), I'm simply not colour-blind enough. Though, most people just don't care of those subtleties...

But quite interestingly, the Lr3 beta defaults do not require any 'calibration' for the a500 files! They look simply perfect in regard of colour rendering. Yet the reds from a500 stay nearly as good as those calibrated from a700, they don't fall down to 'canonesque' appearance at all.

[UrsaMajor]

Does anyone have a solution for this problem? I never experienced it with my earlier Canon DSLRs. Everytime I use the A700 for landscape type shots that have blue sky I realize it has real problems. Back on our long road trip several months ago this problem cropped up all the time.

This may be the same issue that has been discussed several times over the last 6-8 years on the support forum for the image editor that I use. (Picture Window Pro) The most recent discussion of this subject can be found at:

http://www.dl-c.com/board/viewtopic.php?f=4&t=40

My understanding of the discussion at this link is that, as its default the recent versions of Adobe Camera Raw use the ProPhoto color space, which is a very, very wide color space. The author/programmer of Picture Window Pro feels that the use of an extremely wide color space is often a poor choice, as it can force all of the pixels into a small part of the space. This in turn can lead to posterization, especially when the differences in color are subtle - even when processing in 16 bits per channel. He feels that the color space used for processing should be matched to the characteristics of the image file so all of the available colors are included with a minimum of excess - which is why PWP provides a menu with a number of color spaces that makes the choice of color space easy.

You might want to check whether or not this is the reason you are experiencing this problem. Try processing the RAW file again, using a more limited color space such as sRGB or aRGB to see if you continue to have this problem.

With best wishes,
- Tom -

[bakubo]

This looks like a gamut issue to me. There aren't enough colours in the colourspace to implement a stepless blend in the range of tones depicted so the processor is dithering.

Yeah, I shouldn't have used the word noise. The earlier problem in the other thread was about red channel noise in a blue sky. This problem is also in blue skies, but is not noise.

[bakubo]

Your example clearly shows red clipping down in the sky. It's then further exaggerated by 'clarity' setting.

Thanks again for your help. Hmm, I am a bit confused here. When I look at the RGB histogram (you can see it in the screenshot I posted) it shows that none of the channels appear to be clipping and the and the red channel in particular is to the left. I just did another conversion with all values set to 0 and the camera selected white balance of 5400/2. The problem is either gone or much better, but now it has 2 other problems: 1) it is a very dull, uninteresting rendition, 2) the earlier problem of terrible blue sky noise coming from the red channel is prominent, especially when I convert to B&W.

Here's what always works in the 'skies' for me - set recovery much higher at 70 or even more; set blacks to zero; if reds still clip down, add exposure. By all means avoid primaries clipping down in bright saturated areas. After that adjust the tone response by the secondary curve adjustments and brightness/contrast. Only then go to clarity/vibrance/saturation.

Okay, I think I misunderstood above. You meant clipping at the bottom, not the top. I have always tried to not use high settings of recovery because I thought that it was not really such a good thing. It seems at high settings that I often see color shifts, etc. I am definitely learning some new stuff from you. I tried a new conversion using 5400/2 for the color balance, +0.2 for exposure, and 70 for recovery (everything else set to 0). That does even better and there is less of the sky noise that was in the other thread a few months ago and also the current problem is gone. The image is very dull so I will do some post-processing and see what I can do.

I have always thought that it was supposed to be best to get the image as close to what you want during raw conversion while working on the raw data rather than do most of it after the raw conversion. I think what I am getting from you though is that it is better to get a very dull, low contrast, far from finished raw conversion and then do most of the work later. Is that right?

Thanks for having the patience to help. I have been doing this stuff for quite a few years but I will be the first to admit that there are still many things that I don't know that well.

[bakubo]

The ACR always worked best for the Canon. The Canon red response has a wider spectral sensitivity curve, and the red curve also has a long 'shelf' protruding far into the green. This way they avoid red channel noise and down-clipping. But frankly, I cannot stand the 'Canon red' (and the sky-blue also), I'm simply not colour-blind enough. Though, most people just don't care of those subtleties...

Okay, I guess that explains why I never noticed this problem with my Canon DSLRs. Yeah, maybe you see color better than me. I generally didn't have any problem with the look of Canon blue skies. Also, I never had any problem when converting them to B&W. Last spring when I was on the 2 month road trip through the U.S. I ended up with many photos that I converted to B&W and the skies were so horrible I decided to just turn the skies black to eliminate the terrible, blotchy noise that ruined the photos. The black skies gave them an interesting look and I might have gone for it in a few photos anyway, but I felt forced to do it for a bunch because of the red channel noise. If I can get a handle on this problem with the A700 I may go back and reconvert one or two of those old photos to see what happens. Really busy trying to work on my recent Japan photos right now though -- and almost everyday I am adding new ones so it is hard to catch up.

But quite interestingly, the Lr3 beta defaults do not require any 'calibration' for the a500 files! They look simply perfect in regard of colour rendering. Yet the reds from a500 stay nearly as good as those calibrated from a700, they don't fall down to 'canonesque' appearance at all.

I don't suppose you have tried converting any A700 files with the new LR3?

[bakubo]

My understanding of the discussion at this link is that, as its default the recent versions of Adobe Camera Raw use the ProPhoto color space, which is a very, very wide color space. The author/programmer of Picture Window Pro feels that the use of an extremely wide color space is often a poor choice, as it can force all of the pixels into a small part of the space. This in turn can lead to posterization, especially when the differences in color are subtle - even when processing in 16 bits per channel. He feels that the color space used for processing should be matched to the characteristics of the image file so all of the available colors are included with a minimum of excess - which is why PWP provides a menu with a number of color spaces that makes the choice of color space easy.

You might want to check whether or not this is the reason you are experiencing this problem. Try processing the RAW file again, using a more limited color space such as sRGB or aRGB to see if you continue to have this problem.

Since I am processing these photos for my website I am using sRGB.

[agorabasta]

I have always thought that it was supposed to be best to get the image as close to what you want during raw conversion while working on the raw data rather than do most of it after the raw conversion. I think what I am getting from you though is that it is better to get a very dull, low contrast, far from finished raw conversion and then do most of the work later. Is that right?

That's quite right. If you do most adjustments early, you already cut out all the 'unnecessary' raw levels, and hence risk being left with far too reduced data set for a smooth representation in the final image. Exactly that happened in your original image.

Of the use of highlight recovery I should say that at times we absolutely must use it. Deep blue skies with some clouds and whitish horizon have the pure (Sony kind) red to span over all the raw DR of 12bit in a very steep way, i.e. the red raw levels concentrate in the darkest and in the brightest parts of the image and very few are there in midtones. So you really cannot translate it smoothly into 8bit/channel bitmap without DR compression. And if you set blacks above zero for such a scene, the red raw data could be easily reduced to the upper 4bit range, and posterization (bit slicing) is guaranteed.

The camera designers could counter that by letting some 'white' light to all sensels by having the CFA's individual filters' size made smaller than the size of the sensel. Such approach would also lower the noise levels as more light would be caught. But it's done at the cost of poorer colour fidelity. (I strongly suspect Canon is doing exactly that in 7D.)

[agorabasta]

I don't suppose you have tried converting any A700 files with the new LR3?

In fact I did. I find Lr3 the first Adobe product that I really like. And I really mean I like it, not just find it acceptable with reservations. If they don't ruin it on the way to final release, I buy it.

[agorabasta]

Reading the latest review on the new D300s from dpr, the so-called blue sky grain is a by-product of the sensor…

Inasmuch as the image itself is...

[agorabasta]

I presume it’s the same sensor as in the A700, making the problem inherent.

Making so are the laws of nature, the sensor is just delivering it at the eyes of beholder (through his/her electronics).
So you are free to choose whom to blame. The choices are - 1. God/nature; 2. The 'beholder'; 3. The latter's poor devices...