| Author | Message | ||
     Alexander Domagalsky Member Username: Sani Post Number: 26 Registered: 03-2007 |
Hello I have converted raw file to 16bit linear TIFF. In a linear file, dark areas are displayed almost completely black. As soon as gamma 2.2 is applied to linear data I see unpleasant posterization in dark image areas. As I understand it correctly, posterisation is a caused by fact that gamma 2.2 highly increases contrast in shadow areas. My question is: How can I smooth posterized areas? or How can I lighten dark areas so their colors show up? I tried bluring posterized areas but that did not give me pleasing result. Alexander Domagalsky | ||
| Jonathan Sachs Member Username: Jsachs Post Number: 760 Registered: 08-2002 |
This usually happens when areas of a photo are underexposed. Lacking an overexposed version of the same image, I find it is often best just to darken the shadows so as to hide the posterization (which is caused by not having enough information to resolve shadow detail). Jonathan Sachs Digital Light & Color | ||
| Alexander Domagalsky Member Username: Sani Post Number: 27 Registered: 03-2007 |
Jonathan I guess I have to work more on my photographic skills. The camera display showed the brightness histogram of the image in range 15% - 80%, and the widest color histogram was Blue in range 8% - 85%. So it did not look to me like the image was underexposed. Any suggestion how to smooth posterized area. Alexander Domagalsky | ||
| Jonathan Sachs Member Username: Jsachs Post Number: 761 Registered: 08-2002 |
What color management settings are you using? Sometimes using a proofing profile or certain monitor profiles can create posterization. If the histogram was as you describe you should not be seeing much if any posterization. Jonathan Sachs Digital Light & Color | ||
| den Member Username: Den Post Number: 248 Registered: 03-2006 |
Taking a tone range perspective...... This seems to be one of those images… …[a landscape perhaps comprised of distant fine detail that has low contrast, low saturation, and taken at a ISO setting other than 100]… …that needs to retain the maximum available recorded highlight and shadow RGB channel data... at least initially. An approach to achieve this is to specify ‘no expansion’ for contrast when using PWP’s RAW conversion and as the 2.2 gamma seems to be fine for the highlights, continue to use it. For the resulting 16-bit tiff file, perform initial tone changes for black point, white point, brightness, and contrast with the BrightnessCurve transform in the RGB color space with the histogram resolution set to High. One could additionally use a mask to protect the mid-tones and highlights so that more aggressive curve changes could be made to the shadows and lower mid-tones without adversely affecting the upper mid-tones and highlights. Proceed to perform additional preferential brightness, contrast, and color changes in alternate color spaces with the BrightnessCurve or ColorCurve transform with perhaps a final adjustment to the black/white points in the Levels&Color transform with expansion to ‘full range’… …or… Create two RAW conversion files, both with ‘no expansion’ for contrast where a gamma of 2.2 and a higher gamma [shadows are less compressed] are used… Then do a preferential StackImages or Composite-Blend of the resulting images’ 0 to 40% tone ranges… …or… This could very well be one of those images that best presents where the darkest tone is something other than 0%; has only moderate shadow and lower mid-tone contrast and saturation; and has grain [noise] removal. The following link may be of help: http://www.ncplus.net/~birchbay/tutorials/MMs_IMG_0025/RGBcurve/index2-2.htm Do you have a way of making available for download, a full-sized, 100% quality jpeg, and sRGB color image version of your 2.2 gamma conversion tiff file for viewing and experimentation? | ||
| Alexander Domagalsky Member Username: Sani Post Number: 28 Registered: 03-2007 |
Jonathan Posterization is visible even if I convert my raw file to 16bit TIFF with the tool Canon Raw Image Task. So the problem is not in PWP raw conversion. In window "Color Management Settings" I have parameter Profing Profile set to None. I use common/office monitor BenQ FP992. You think that the posterization is the result of the non-graphic monitor but the data in the file are OK? You might be really right, when I calibrated my monitor I noticed that LUT curves close to point [0, 0] were not very linear. Alexander Domagalsky | ||
| Alexander Domagalsky Member Username: Sani Post Number: 29 Registered: 03-2007 |
Hello Den Thanks for all your suggestion. I will try it all. The image is quite simple here is just its small version. I do not have way to provide image for download, but I can sent the full size version to you by email.  Alexander Domagalsky | ||
| Jonathan Sachs Member Username: Jsachs Post Number: 762 Registered: 08-2002 |
Another possibility - this may be your camera's noise reduction algorithm at work. If you can, try turning it off or reducing it. Jonathan Sachs Digital Light & Color | ||
| den Member Username: Den Post Number: 249 Registered: 03-2006 |
JS - good point, as I prefer to do noise reduction in post-processing, in-camera settings would not have occurred to me. I usually just try to use the lowest possible ISO setting that will still give me the 'capture' to control noise in the camera's download image. AlexanderD: There is no need to email an image. I have plenty of similar images. Your concerns seem to be typical renderings of digital cameras for soft focus, low contrast image areas of this type, at least for my cameras. My typical post-processing approach is as JS originally suggested, i.e., darken the background with preference noise reduction... A possible simplified post-processing approach to lighten and de-noise the background could be: 1. Use MaciejT's 'contrast mask' technique to adjust background and main subject image areas to preference. From the posted image, it appears that the contrast mask's 50% tone should be around the image's 38% tone. 2. Use the Step1 contrast mask and AdvancedSharpen - Noise Reduction to reduce the background noise to preference. [A personal prefence of mine is to sometimes over-do the noise removal, then using the AddNoise transform, add back 1 to 3% gaussian RGB noise.] 3. Use the Step1 contrast mask and AdvancedSharpen - Sharpen to sharpen to the main subject to preference.  | ||
| Alexander Domagalsky Member Username: Sani Post Number: 30 Registered: 03-2007 |
Thanks Jonathan and Den I will play with my image to use suggested techniques. Alexander Domagalsky | ||
| Alexander Domagalsky Member Username: Sani Post Number: 31 Registered: 03-2007 |
Den I have one question for you. In your suggested workflow w/ MaciejT's contrast mask you said: "it appears that the contrast mask's 50% tone should be around the image's 38% tone" What does that mean? How do you suggest to modify the simple MacejT's mask brightness curve [30, 0], [70, 100] to take into consideration 38% in my image? Alexander Domagalsky | ||
| den Member Username: Den Post Number: 250 Registered: 03-2006 |
Open the Mask Tool - Brightness Curve and set histogram resolution to High from the OPT drop downs... MaciejT's suggested curve with a 40/100 slope is shown in yellow... set a BrokenLine curve so that it has a 40/100 slope that intersects [38,50]... note that the lowest histogram dip between the 'two peaks' occurs at the 38% tone...  If maximum separation of the flower and the background is desired, try a StepCurve at the 38% tone...  For my illustration, I used a Mask Blur=10 for the 600 pixel wide image you posted... for the full sized image, you might start by linearly ratio-ing this blur upward by the wide pixel dimension of the full sized image,i.e., Blur = 10 x (FS wide pixels/600). Comments: The secret to a 'contrast mask' method is to properly locate the mask's 50% tone so that it coinsides with the image's tone value that you want to remain unchanged... for two peak histogram images, this is usually the lowest dip between the two peaks... Notice too how easily tone range/histogram considerations can be used to surgically mask image content! Hope this is of help... | ||
| Alexander Domagalsky Member Username: Sani Post Number: 32 Registered: 03-2007 |
Hello Den This is absolutelly fantastic explanation. Thank you very much. Alexander Domagalsky | ||
| den Member Username: Den Post Number: 251 Registered: 03-2006 |
You are welcomed. It seems we may live within an hour's drive from each other, not counting border crossing delays... Contact me if there is an interest. | ||
| MJayne Member Username: Mjayne Post Number: 67 Registered: 10-2006 |
Hi, I have a question about this workflow shown here: Why is the yellow curve line called a 40/100 curve if you go over 3 on the botton and 3 on the top? | ||
| den Member Username: Den Post Number: 262 Registered: 03-2006 |
The 'slope 40/100' refers to the change in horizontal coordinates divided by the change in vertical coordinates of curve control points 2 and 3....  | ||
| Bob Walker Member Username: Rwalkernm Post Number: 131 Registered: 06-2004 |
MJayne, The curve goes from 0 to 100 (bottom to top) over a run of 40 units from left to right. It's positionable anywhere along the left to right, and in den's example, he puts the place where the curve crosses 50 (bottom to top) where there is a minimum in the histogram. Bob | ||
| MJayne Member Username: Mjayne Post Number: 69 Registered: 10-2006 |
Thanks Guys, I should have known that. I guess I was just looking at the two points, 30,0 and 70,100, rather than looking at it from a rise over run perspective. |
