Results 1 to 38 of 38

Thread: Processing raw image exercise for February 2012

  1. #1
    BPN Member Roger Clark's Avatar
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,956
    Threads
    257
    Thank You Posts

    Default Processing raw image exercise for February 2012

    Attached Images Attached Images
     
    The first raw processing image for February comes from Mital Patel.

    The raw image in dng format can be downloaded from here:
    http://www.clarkvision.com/bpnraw/mi...l.img_9901.dng

    The image is well focused, sharp, underexposed by only about 1/2 stop. The challenges as I see it is the bird is small in the frame, so after cropping, noise, and detail will be a challenge to control. Submit your best efforts in replies to this thread with maximum dimensions 1024 wide, by 800 high, sRGB jpegs. The file size limit is currently 256 KBytes but I'll try and get that raised to 350 KB like in the critique forums.

    With a good response, we'll do two per month if people do not get tired out. You are also welcome to try your processing on previous month's exercises too.

    Roger

  2. #2
    BPN Member Tom Graham's Avatar
    Join Date
    Apr 2010
    Location
    Southern California, Orange County
    Posts
    1,114
    Threads
    33
    Thank You Posts

    Default

    I'll have a go. My 20mins in Photoshop and Noise Ninja. Severly cropped to only about 25% of original pixels retained.

    Name:  mital.patel.img_9901 B.jpg
Views: 419
Size:  126.0 KB

    (Yes, I flipped it left/right).
    As I see it more, I wonder if the water should be more saturated color? But that might distract from the little color in the bird?
    Hmmm, is water too cyan??
    Tom
    and many thanks Roger for your work/help with us on such exercises

  3. #3
    Forum Participant John Chardine's Avatar
    Join Date
    Jan 2008
    Location
    Canada
    Posts
    6,631
    Threads
    650
    Thank You Posts

    Default

    Hi Roger- Tried the download using latest Safari and Lion and all I get is garbage text up on the screen. Previous download (egret) works great with same system. Any thoughts?

    Update- right-clicked link and chose download linked file and it seems to be working. I'll know more after the download.

  4. #4
    BPN Member Jerry van Dijk's Avatar
    Join Date
    Sep 2008
    Location
    The Netherlands
    Posts
    1,686
    Threads
    162
    Thank You Posts

    Default

    Attached Images Attached Images
     
    Here's my go at it. Cudos for catching this speed devil at this quality!
    Main work in ACR 6.3:
    16:9 crop
    WB adjustment (Daylight preset: 5500, +10)
    Exp: +0.25
    Clarity: +59
    Vibrance +37
    Curves adjustment: Black point to 29, white point to 231, reverse S-curve

    Sharpness:
    Amount 48
    Radius 0.8
    Detail 38
    Masking 31
    NR
    Luminance 6
    luminance detail 50
    Color 25
    Color detail 50

    Further processing in CS5.
    Cropped from the right
    Removed some debris from water using clone stamp
    Selective NR on BG
    Selective USM on bird
    USM on whole image
    Increased contrast and sharpness (USM) on eye
    Save for web and devices

  5. #5
    BPN Member Roger Clark's Avatar
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,956
    Threads
    257
    Thank You Posts

    Default

    Great job so far. Keep them coming.
    Roger

  6. #6
    Forum Participant Mital Patel's Avatar
    Join Date
    May 2009
    Location
    Ahmedabad, Gujarat, INDIA
    Posts
    2,061
    Threads
    319
    Thank You Posts

    Default

    Attached Images Attached Images
     
    Here's Mine

    Open in ACR CS5


    Camera Standard Profile
    Temperature : 4900
    Tint : +10


    Exposure +60
    Recovery 18
    Fill Light 15
    Blacks 8
    Brightness +50
    Contrast +25
    Vibrance +15


    Strong Contrast


    Detail Tab
    Sharpening 25, 1.0, 25, 10
    Noise Reduction 25, 50, 0, 60, 60


    OPE IMAGE
    Cropped for size


    Masked bird
    Applied Noise Reduction bg
    Default : 6,15,19,8


    USM 180,0.3,0


    Save As JPG

  7. #7
    Forum Participant John Chardine's Avatar
    Join Date
    Jan 2008
    Location
    Canada
    Posts
    6,631
    Threads
    650
    Thank You Posts

    Default

    Attached Images Attached Images
     
    How can you argue with the artist himself?! Here's my "go". In ACR I bumped default exposure to +1.00, Blacks from 5-10 and ran some capture sharpening and chroma NR before bringing into Ps. I cropped a little tighter than above, because the image was able to take the crop and I wanted the subject to have more of an impact. I ran Topaz Denoise on the BG, increased the saturation in LAB A and B channels one "notch", and sharpened the subject.

    Up close the bird is very "ratty" and needs a good moult!

    Tom- why did you turn the bird around?

  8. #8
    Forum Participant
    Join Date
    Oct 2009
    Posts
    119
    Threads
    48
    Thank You Posts

    Default

    Here's the result of some quick work in Lightroom. I tried to keep more of the background and not make the bird quite so big, especially since the head doesn't have a lot of detail in it. This source seems to work much better as a bird behavior photo (small swallow skimming along the water) than as a bird portrait.

  9. #9
    BPN Member Roger Clark's Avatar
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,956
    Threads
    257
    Thank You Posts

    Default

    Attached Images Attached Images
     
    Here is my go at the processing:

    open in photoshop acr

    Noise reduction:
    luminance 25
    luminance detail 50
    luminance contrast 0
    color 50
    color detail 65

    white balance as shot

    exposure 0
    recovery 23
    fill light 0
    blacks 0
    brightness +50
    contrast +25
    clarity +8
    vibrance +9
    saturation +12

    crop

    In photoshop:

    levels right slider to 230

    select dark parts of bird with magic wand tool
    refine edge:
    edge detection 0.9 px
    smooth 12
    feather 20
    contrast 8
    shift edge 8

    curves:
    in= 192 out= 192
    in= 57 out= 75
    deselect

    curves on whole image:
    in= 188 out= 195
    in= 102 out= 103
    in= 37 out= 24

    save as version b.

    in ImagesPlus:
    Richardson-Lucy deconvolution: 5x5, 25 iterations

    Back to photoshop

    Bring in new Richardson-Lucy deconvolution as a layer on top of previous version.
    select background and delete (noise was enhanced in the deconvolution)

    some further selective noise reduction on background (Gaussian blur 0.8)

    merge layers.

    select back of bird with lasso tool
    edge detection 0.9 px
    smooth 12
    feather 20
    contrast 8
    shift edge 8
    curves tool to lighten back of bird
    in= 113 out= 117
    in= 47 out= 57
    in= 16 out- 18

    curves tool to bring out feather detail
    in= 126 out= 129
    in= 71 out= 74
    in= 38 out= 30

    some general (minor) clean up (blur tool) on some edges that had residual ringing from processing.

    convert to sRGB, 8-bit,
    size for web and save for web

    Roger

  10. #10
    BPN Member Tom Graham's Avatar
    Join Date
    Apr 2010
    Location
    Southern California, Orange County
    Posts
    1,114
    Threads
    33
    Thank You Posts

    Default

    Roger's is the clear winner.
    Mine is definitely much too cyan all over. (I must remember to leave an image overnight then review again before posting!!)
    "Tom- why did you turn the bird around?" I did that because I feel sometimes it gives the feeling of the subject entering/flying into the image rather than out. For same reason I generally prefer a main light coming from the left rather than right. In this image both subject and light were right. Why left not right? I've heard it is a cultural (western?) habit and associated perhaps with our reading left to right. And, FWIW, for example, Rembrandt is well known for his strong light coming from the left.
    Tom

  11. #11
    Lifetime Member Michael Gerald-Yamasaki's Avatar
    Join Date
    Jan 2010
    Location
    Santa Cruz, CA USA
    Posts
    1,429
    Threads
    213
    Thank You Posts

    Default

    Folks,

    Greetings. I tend to like images on the warmer side as a matter of taste. Like the orange-blue color contrast in the image and prefer the yellow orange side to the red orange side of the palette for this image.

    Name:  mital.patel.img_9901-2.jpg
Views: 291
Size:  158.1 KB

    ACR
    WB 5800 Tint +8

    Exposure +2.0
    Blacks +10
    Brightness 0 /prefer to take brightness in exposure, but am a novice in ACR
    Contrast 0

    Clarity +10
    Vibrance +10

    Convert to Prophoto for work in PS

    Mostly Topaz for grins

    Topaz Denoise - RAW light w/ Adjust Highlight to 0.2 whole image
    High Pass Sharpening w/Hard Light Blend - been playing with this again lately for detailing includes sharpening the high pass layer
    Topaz Simplify - Buz Sim preset with Simplify Size 0.01 Saturation Boost 1.00 - Essentially NR/Detailing/Contrast/Saturation in Topaz's segmented style - here for color
    Topaz InFocus - for detailing
    Topaz Denoise - Selective for bg RAW Moderate settings

    crop & save to tiff

    jpeg w/standard screen sharpening in LR

    Cheers,

    -Michael-

  12. #12
    Lifetime Member Michael Gerald-Yamasaki's Avatar
    Join Date
    Jan 2010
    Location
    Santa Cruz, CA USA
    Posts
    1,429
    Threads
    213
    Thank You Posts

    Default

    Personally, I would like to see more comments about the various efforts here. So, here are mine:

    Tom - After working on the image the flip just seems strange though I understand the directionality preference. Agree with the color has a cyan edge comment. BG seems a bit noisy. Could be sharper. Good tone & saturation. With some minor color adjustment & detailing...

    Jerry - A little dark for my taste. Why the reverse-s curve (flattens midtones, not enough whites to matter)?

    Mital - Too cool for me.

    John - I think we went in a similar direction. I like the color. Prefer some additional detailing. I think the blue edges on the top wing are a problem (from boosting darks).

    Elliotte - There is some appeal to the larger view. The warmth is a bit too far for me (water is no longer predominantly blue). Could use a bit more detail.

    Roger - Good detail though a little crispy in spots. Problem blue edges in the darks (to my eye), perhaps lifted a bit too much for my taste.

    Mine - Could probably use a bit more room on the right. A bit too crispy in places. Lesser, but still a problem with the blue edges on the top wing. Nice color

    JMO...

    Cheers,

    -Michael-

  13. #13
    BPN Member Tom Graham's Avatar
    Join Date
    Apr 2010
    Location
    Southern California, Orange County
    Posts
    1,114
    Threads
    33
    Thank You Posts

    Default

    Hey Michael - agree with you, not enough comments and critique. (Also true of other forums, too much - "I like it", with no reasons). So interesting to have yours on each. The bird is of such strange/unfamiliar color for me that it was difficult to decide what is close to real life. Although I admit for all images I lean towards "make it look good".
    Flipping image goes way back for me when I had a 35mm chrome landscape with a path going foreground off into distance. The original (as taken) had path curving in from the right corner. When I put it into club competion the judges told me it would score higher if I would flip it so path lead in from left.
    Tom

  14. #14
    BPN Member Roger Clark's Avatar
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,956
    Threads
    257
    Thank You Posts

    Default

    A couple of comments on the processing.

    Be careful with the use of the exposure slider. The exposure slider in ACR is like increasing the exposure in the camera: it acts linearly on the data. The brightness slider increases brightness after the gamma curve is applied. The difference is that it is very easy to drive the highlights into saturation with the exposure slider. The brightness slider is more forgiving. Michael's image, where he increased the exposure to +2 has saturated parts of the image in some channels. The two areas where this is most apparent are the bright areas below the eye and the small white tail feathers. Check, for example, the red channel below the eye, and one will see a lack of detail. Bring the images posted here into photoshop and put small selection boxed around these areas and look at the (all 3 colors) histgram. This will show if some data are hitting the limits.

    One does not have the tools in ACR to be certain if a small area is getting a channel saturated, so I prefer to be conservative in ACR and do the adjustments in photoshop where I have more tools to check these things.

    Similarly, the black slider: It is too easy to clip the lows, so I set blacks to zero in ACR. If any data gets clipped during the raw conversion, information is lost for the rest of processing. In Michael's image, some areas on the back of the wing are clipped to zero in some channels and in some locations, all 3 channels are clipped. Michael, I'm not trying to pick on you, as I like your overall processing, just do not like clipped highs or lows, and I think it is important to point out the dangers of sliders that can clip data.

    Roger
    Last edited by Roger Clark; 02-26-2012 at 10:27 AM.

  15. #15
    BPN Member Tom Graham's Avatar
    Join Date
    Apr 2010
    Location
    Southern California, Orange County
    Posts
    1,114
    Threads
    33
    Thank You Posts

    Default

    Regarding finding in an image areas of over saturation, 255, or clipped lows, 0. Must be an easier way to find all such image areas in Photoshop other than using the eyedropper and manually scanning the image with it while watching for 255s or 0s readings?
    Tom

  16. #16
    BPN Member Roger Clark's Avatar
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,956
    Threads
    257
    Thank You Posts

    Default

    Quote Originally Posted by Tom Graham View Post
    Regarding finding in an image areas of over saturation, 255, or clipped lows, 0. Must be an easier way to find all such image areas in Photoshop other than using the eyedropper and manually scanning the image with it while watching for 255s or 0s readings?
    Tom
    Hi Tom,

    The way to find clipped areas is to use the levels tool. Hold down the alt key (on windows) and move either the right or left sliders. The image will go black but as you move the slider away from the limit, if any data are clipped it will show as a color. The color and intensity depends on how many channels and how much it is clipped.

    Roger

  17. #17
    Lifetime Member Michael Gerald-Yamasaki's Avatar
    Join Date
    Jan 2010
    Location
    Santa Cruz, CA USA
    Posts
    1,429
    Threads
    213
    Thank You Posts

    Default

    Roger,

    Greetings. Thank you very much. This is exactly the type of comment I was hoping for from this forum. I particularly like the explanation of exposure vs brightness. I've never particularly liked brightness adjustments largely because different tonal areas respond differently to the adjustments. Now, I understand it is the gamma curve which is moderating the adjustments in different tonal areas. Ah, hah! Now I know why I like exposure adjustment better, for me the effect of the adjustment as a tool is just more predictable than the brightness adjustment. Or another way of putting it is I would rather take the overall decrease in contrast from a linear adjustment than the flattening of darks and highlights (no matter the image) that the gamma curve imposes.

    Blown single channels in whites in small areas (no obvious banding) doesn't bother me. Under the eye is a bit more of a concern since the blown channel is the predominant color but here I'm not bothered. (especially as it is an expression of a specular highlight where there would be reduced contrast of any sort ). Most of the blown parts btw came from the +10 in vibrance that I applied after the exposure.Similarly (perhaps, not surprisingly), I don't have much issue with clipping small areas to black.

    Since the original capture was underexposed everything has to be boosted w/ brightness, exposure or other means. Boosting darks introduces a problem for color in all cases. All color models show decreased color resolution in the darks. Digitized color models further reduces color resolution. So, when brightening darks, the low resolution in color is expressed at a brightness level that is capable of greater resolution. Visually the colors drift towards primarys (with respect to full color resolution). This effect is illustrated in (all) these images by the blue in the darks of the back wing (not the shoulder layer, but the next layer of feathers) while it is not in evidence in the front wing which is better exposed. The impact of reduced color resolution in darks effectively places a limit on how much they can be boosted, while maintaining "realistic" color balances. How one chooses to render even use the impact of reduced color resolution in darks is of course a matter of taste.

    Btw, there is a similar reduction of color resolution in the whites (!), but there weren't any in this image due to underexposure. So, color issues associated with toning down whites (or recovery) aren't in evidence for this image due to underexposure.

    Cheers,

    -Michael-
    Last edited by Michael Gerald-Yamasaki; 02-26-2012 at 02:08 PM. Reason: typo

  18. #18
    BPN Member Roger Clark's Avatar
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,956
    Threads
    257
    Thank You Posts

    Default

    Quote Originally Posted by Michael Gerald-Yamasaki View Post
    Roger,

    Since the original capture was underexposed everything has to be boosted w/ brightness, exposure or other means. Boosting darks introduces a problem for color in all cases. All color models show decreased color resolution in the darks. Digitized color models further reduces color resolution. So, when brightening darks, the low resolution in color is expressed at a brightness level that is capable of greater resolution. Visually the colors drift towards primarys (with respect to full color resolution). This effect is illustrated in (all) these images by the blue in the darks of the back wing (not the shoulder layer, but the next layer of feathers) while it is not in evidence in the front wing which is better exposed. The impact of reduced color resolution in darks effectively places a limit on how much they can be boosted, while maintaining "realistic" color balances. How one chooses to render even use the impact of reduced color resolution in darks is of course a matter of taste.

    Btw, there is a similar reduction of color resolution in the whites (!), but there weren't any in this image due to underexposure. So, color issues associated with toning down whites (or recovery) aren't in evidence for this image due to underexposure.

    Michael,
    I'm not sure I understand what you mean with your above statement. For example, what do you mean by "Digitized color models further reduces color resolution." In a digital camera image, the low level signals are pretty linear, even after application of the variable gamma curve (the standard characteristic curve). What I see when lifting shadows is posterization. There is another issue with some raw converters that drop color in shadows. ACR was not great back in CS2 days:
    http://www.clarkvision.com/articles/...shadow.detail/
    I haven't updated that page in a while, but I have tried a more recent ACR with similarly poor results (think it was CS4, but maybe it was CS3).
    Is any of this what your are referring to? If so, this ia purely processing algorithm problems and not anything inherent in the digital data.

    Regarding loss of color in the high end, there is loss of color and tonality (posterization) due to the characteristic curve which compresses the data into a smaller range, losing fine tonality levels. If one needs to get more detail (color and/or tonality) in highlights, do a linear conversion (which photoshop's ACR will not do).

    Roger

  19. #19
    Lifetime Member Michael Gerald-Yamasaki's Avatar
    Join Date
    Jan 2010
    Location
    Santa Cruz, CA USA
    Posts
    1,429
    Threads
    213
    Thank You Posts

    Default

    Quote Originally Posted by Roger Clark View Post
    The way to find clipped areas is to use the levels tool. Hold down the alt key (on windows) and move either the right or left sliders. The image will go black but as you move the slider away from the limit, if any data are clipped it will show as a color. The color and intensity depends on how many channels and how much it is clipped.
    Curves also works this way with the right and left sliders.

    (Just to be clear) Leaving the slider at 0 or 255 shows you what you have, moving them shows you what you will have if you make that adjustment. The color is which channels are at 255 or 0 for each pixel (for the right hand slider, for instance, white for all three channels, yellow for red and green channels, etc.).

    Cheers,

    -Michael-

  20. #20
    BPN Member Roger Clark's Avatar
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,956
    Threads
    257
    Thank You Posts

    Default

    I just check and did some new processing tests for my page
    http://www.clarkvision.com/articles/...shadow.detail/

    I previously checked CS3, and no improvement over what is on the page. So I just tested CS5 and it is a small improvement over that from CS2 in figure 8, but only a little, and nowhere near as good as ImagesPlus 2.5. Dcraw results are really close to that of CS5 ACR 6.6

    Roger

  21. #21
    Lifetime Member Michael Gerald-Yamasaki's Avatar
    Join Date
    Jan 2010
    Location
    Santa Cruz, CA USA
    Posts
    1,429
    Threads
    213
    Thank You Posts

    Default

    Quote Originally Posted by Roger Clark View Post
    Michael,
    I'm not sure I understand what you mean with your above statement. For example, what do you mean by "Digitized color models further reduces color resolution." In a digital camera image, the low level signals are pretty linear, even after application of the variable gamma curve (the standard characteristic curve). What I see when lifting shadows is posterization. There is another issue with some raw converters that drop color in shadows. ACR was not great back in CS2 days:
    http://www.clarkvision.com/articles/...shadow.detail/
    I haven't updated that page in a while, but I have tried a more recent ACR with similarly poor results (think it was CS4, but maybe it was CS3).
    Is any of this what your are referring to? If so, this ia purely processing algorithm problems and not anything inherent in the digital data.

    Regarding loss of color in the high end, there is loss of color and tonality (posterization) due to the characteristic curve which compresses the data into a smaller range, losing fine tonality levels. If one needs to get more detail (color and/or tonality) in highlights, do a linear conversion (which photoshop's ACR will not do).

    Roger
    Roger,

    What I mean about color resolution in digital color:

    For simplicy, let's look at the color representation space for 8 bit rgb. In Cartesian space, it could be represented as a cube 0 to 255 on each edge. The diagonal of the cube would be 0,0,0 to 255,255,255 and also represent a gray scale from 0 to 255. If the diagonal is oriented horizontally, this would correspond more or less to 0-255 in the curves or histograms we are used to looking at. The point of this visual image (a cube lying on it's diagonal) is to get a sense of how many colors (hues) there are at each level of brightness from 0 to 255. Since the shape of the color space (a cube) comes to a point at black and white, the number of hues in the darks and in the whites are small in comparison to brightness levels as they approach midtones. This isn't a problem until you start modifying colors (brightening darks or darkening whites).

    Let me give an example:

    Say you have a set of pixels at or below gray level 1 (just above black). There are only 7 possible hues including neutral, 8 total colors: neutral (0,0,0 and 1,1,1), red (1,0,0), green(0,1,0), blue (0,0,1), yellow (1,1,0), cyan (0,1,1), and magenta (1,0,1). This is the impact of digital color: you can't represent a very dark burnt orange for instance. But it all doesn't matter because at this gray level everything pretty much looks black anyway. Unless...

    ... one boosts the dark into a higher level. Say in this example we boost the one of these pixels to level 50. You still only have 7 hues. This is where posterization would show the low resolution in the digitized color. Without the ability to represent burnt orange at level 1, it certainly won't show up at level 50 where it can be seen.

    I hope that's clear ... the impact of the low resolution of digital color in the darks in the face of boosting. Whites work just the opposite (replace 0,0,0 with 255,255,255 and 1,0,0 [red] with 254,255,255 [cyan]... etc. in the above example. The issue is the same low resolution of digital color in the face of recovery/darkening.

    Regarding linear conversion. Nikon's Capture NX2, the converter that I use has a conversion curve editor (they call it Picture Control Utility) that I use from time to time to make custom conversion curves (you can even load these curves into your Nikon camera). I like this feature of CNX a lot. Nikon's have a standard Neutral Picture Control. Whether this is truly neutral or not I've never checked.

    Cheers,

    -Michael-

  22. #22
    BPN Member Roger Clark's Avatar
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,956
    Threads
    257
    Thank You Posts

    Default

    Quote Originally Posted by Michael Gerald-Yamasaki View Post
    Roger,

    What I mean about color resolution in digital color:

    For simplicy, let's look at the color representation space for 8 bit rgb. In Cartesian space, it could be represented as a cube 0 to 255 on each edge. The diagonal of the cube would be 0,0,0 to 255,255,255 and also represent a gray scale from 0 to 255. ...
    Michael,
    Your example is fine except for one thing. What you are describing is simple posterization. I did all my processing in photoshop 16-bit mode, so pesterization should be negligible. The camera output is 14-bit, which, when combined with photon and read noise should also make the images immune from such posterization effects that we can see. So your example gets turned into 0 to about 3800 out of camera (real values from the camera), which is well below the tonality that can be seen with the human eye. Then gets turned into 0 to 32765 levels for "16-bit" photoshop, again reducing any possibility of visible posterization. The problem we see seems to come from photoshop's relatively poor raw conversion algorithm, as illustrated on the page I referenced.

    So, for example, your 7 levels turns into 7*(3800/255=) 104 levels out of camera, which is plenty, and that then gets converted to about 896 levels (104*32767/3800) in the 16-but tiff file. So posterization from the camera or post raw conversion is not a factor.

    Roger
    Last edited by Roger Clark; 02-26-2012 at 06:50 PM.

  23. #23
    BPN Member Tom Graham's Avatar
    Join Date
    Apr 2010
    Location
    Southern California, Orange County
    Posts
    1,114
    Threads
    33
    Thank You Posts

    Default

    Thanks guys, interesting discussion. A real learning opportunity for me .
    Tom

  24. #24
    Lifetime Member Michael Gerald-Yamasaki's Avatar
    Join Date
    Jan 2010
    Location
    Santa Cruz, CA USA
    Posts
    1,429
    Threads
    213
    Thank You Posts

    Default

    Quote Originally Posted by Roger Clark View Post
    Michael,
    Your example is fine except for one thing. What you are describing is simple posterization. I did all my processing in photoshop 16-bit mode, so pesterization should be negligible. The camera output is 14-bit, which, when combined with photon and read noise should also make the images immune from such posterization effects that we can see. So your example gets turned into 0 to about 3800 out of camera (real values from the camera), which is well below the tonality that can be seen with the human eye. Then gets turned into 0 to 32765 levels for "16-bit" photoshop, again reducing any possibility of visible posterization. The problem we see seems to come from photoshop's relatively poor raw conversion algorithm, as illustrated on the page I referenced.

    So, for example, your 7 levels turns into 7*(3800/255=) 104 levels out of camera, which is plenty, and that then gets converted to about 896 levels (104*32767/3800) in the 16-but tiff file. So posterization from the camera or post raw conversion is not a factor.

    Roger
    Okay, so I over simplified ... Let me try again. The issue isn't posterization so the above discussion while true isn't completely relevant (largely my fault for introducing posterization), but let's put that aside for a moment.

    Most color models have a characteristic shape:

    Name:  ColorModels.jpg
Views: 273
Size:  27.6 KB

    RGB will be a cube with the diagonal in the middle as discussed before. If you look at 3D gamut maps, you will see something like a distorted cube but still with a narrowing of the shape at the bottom (darks) and the top (whites). The width of the 3D shape is at any given brightness level is representative of color resolution, that is, how many colors can be represented at that level of brightness. It's not an issue that there is lower resolution in the darks and whites because one can't distinguish between dark or white colors to any high degree of resolution.

    All this is fine and good until one wants to boost a color from one level to another. The question is what values are assigned to the boosted color? Does (10, 5, 0) become (50, 25,0) or (50, 45, 40)? or something in between. Is there a linear boost or is it controlled by some curve adjustment for the whole color space? What impact does white balance have? Is this being done in conversion or post-processing? We discussed how brightness and exposure express boosting differently. An additional issue for darks is what noise reduction is in play and how much color noise is there?

    In any event, all I'm saying is boosting levels injects artifacts in the color. Sometimes the artifacts are more noticeable than others. In this image set the blue in the back wing feathers is a problem to my eye. This is irrespective of one doing the boosting in 16-bit ProPhoto, though it will be a smoother artifact than for 8-bit sRGB.

    Another way of saying it is the color accuracy isn't very good in darks because of low color resolution, but that's okay because, well, the color is dark... er, until it isn't.


    Cheers,

    -Michael-

  25. #25
    BPN Member Roger Clark's Avatar
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,956
    Threads
    257
    Thank You Posts

    Default

    Quote Originally Posted by Michael Gerald-Yamasaki View Post

    In any event, all I'm saying is boosting levels injects artifacts in the color. Sometimes the artifacts are more noticeable than others. In this image set the blue in the back wing feathers is a problem to my eye. This is irrespective of one doing the boosting in 16-bit ProPhoto, though it will be a smoother artifact than for 8-bit sRGB.

    Another way of saying it is the color accuracy isn't very good in darks because of low color resolution, but that's okay because, well, the color is dark... er, until it isn't.
    Michael,

    I see it differently. There are two effects that impact the image quality. With 16-bit processing, there is plenty of precision for what you talk about (quantization) to not be an issue. But what is an issue is noise. When one boosts shadows, there is noise in each channel, and in shadows/low signal areas, that is a combination of photon noise, read noise and the downstream electronics. Boosting the shadows/dark areas boosts the noise too, and with the lower signal-to-noise ratios of low signals, it can look pretty bad. Add to that fact is processing. I agree that processing can induce color shifts, unless one works in lab mode on the luminance channel (which I often do but did not with this image).

    In the case of this raw processing thread, it would be nice to hear from Mital to hear what color the bird really was, and if any processing color shifts are objectionable.

    Roger

  26. #26
    Lifetime Member Michael Gerald-Yamasaki's Avatar
    Join Date
    Jan 2010
    Location
    Santa Cruz, CA USA
    Posts
    1,429
    Threads
    213
    Thank You Posts

    Default

    Quote Originally Posted by Roger Clark View Post
    Michael,

    I see it differently. There are two effects that impact the image quality. With 16-bit processing, there is plenty of precision for what you talk about (quantization) to not be an issue. But what is an issue is noise. When one boosts shadows, there is noise in each channel, and in shadows/low signal areas, that is a combination of photon noise, read noise and the downstream electronics. Boosting the shadows/dark areas boosts the noise too, and with the lower signal-to-noise ratios of low signals, it can look pretty bad. Add to that fact is processing. I agree that processing can induce color shifts, unless one works in lab mode on the luminance channel (which I often do but did not with this image).

    In the case of this raw processing thread, it would be nice to hear from Mital to hear what color the bird really was, and if any processing color shifts are objectionable.

    Roger
    Roger,

    I'm not talking about precision or quantization. But we are in agreement about noise. I did a neutral as possible conversion to look at the color in an unboosted version (no brightness or exposure adjustments). For color in the darks in this image, I'm afraid noise dominates.

    Looking at it again I continued from the completely neutral conversion and applied a more normal workflow for me in LAB curves. More or less matched the tonal features of my first effort but (iMO) corrected the color in the dark back wing.

    Name:  mital.patel.img_9901-Edit.jpg
Views: 264
Size:  209.0 KB

    This is about what would be a final for me (missed a little nr in the reflection, perhaps a tad more sharpening).

    Cheers,

    -Michael-

  27. #27
    BPN Member Tom Graham's Avatar
    Join Date
    Apr 2010
    Location
    Southern California, Orange County
    Posts
    1,114
    Threads
    33
    Thank You Posts

    Default

    FWIW, I tried comparing the works by Roger (#9) and Michael (#26). Cropped into head body area. Had to do some resizing which changed the image detail a bit, here is result -
    Name:  RCMGY1.jpg
Views: 262
Size:  108.6 KB

    Tom
    Last edited by Tom Graham; 02-27-2012 at 01:48 PM. Reason: added #9, 26

  28. #28
    Lifetime Member Michael Gerald-Yamasaki's Avatar
    Join Date
    Jan 2010
    Location
    Santa Cruz, CA USA
    Posts
    1,429
    Threads
    213
    Thank You Posts

    Default

    Quote Originally Posted by Tom Graham View Post
    FWIW, I tried comparing the works by Roger (#9) and Michael (#26). Cropped into head body area. Had to do some resizing which changed the image detail a bit, here is result -
    Tom, You missed the part that is most at issue which is the back wing.

    Cheers,

    -Michael-

  29. #29
    BPN Member Roger Clark's Avatar
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,956
    Threads
    257
    Thank You Posts

    Default

    Attached Images Attached Images
     
    Quote Originally Posted by Michael Gerald-Yamasaki View Post
    Tom, You missed the part that is most at issue which is the back wing.
    This gets into subjective areas. The light from the sun is low behind the photographer. The far wing is tilted away from us, so is receiving less sunlight. Thus, that wing has more relative illumination from the blue sky, and therefore even if the color was neutral would show in the photo as blue. But the wing feathers do not look black to me. The close wing shows beautiful iridescent blue/blue-green color. Those colors are certainly on the far wing too. So in my opinion if the close wing is not clearly black, neither should the far wing appear black.

    Then one gets into interpretation of color cast due to different color lighting (e.g. sun and sky in this case). I would err on the side of keep the lighting natural light the way it existed on site as opposed to correct to "true reflectance" of the subject. I attach an example of this effect with sandhill cranes at Bosque: the cranes have gray wings, and this is a golden sunset image, but the blue sky is reflecting off of the wings, making them appear blue. I like the effect. Similarly, I would not change the color of the wing of the bird in this raw processing thread, but that is my interpretation and there is no right or wrong answer in my opinion. I chose to keep the blue on the far wing. One can next argue about is the blue the correct shade of blue? Probably it has been modified through my processing with curves.

    Roger

  30. #30
    BPN Member Tom Graham's Avatar
    Join Date
    Apr 2010
    Location
    Southern California, Orange County
    Posts
    1,114
    Threads
    33
    Thank You Posts

    Default

    Back up in #13 I said I had no clue as to real color of this bird. And if it is iridescent, then a couple degrees shift of view angle is going to make big color shifts. Obviously then a very difficult moving subject to photograph. As seen real life, our eye-brain will add it all up and compose an image that is not in any one single photo. And very pretty photo Roger.
    Tom

  31. #31
    Lifetime Member Michael Gerald-Yamasaki's Avatar
    Join Date
    Jan 2010
    Location
    Santa Cruz, CA USA
    Posts
    1,429
    Threads
    213
    Thank You Posts

    Default

    Roger,

    I appreciate the discussion. My original comments on boosting darks were largely to say that there's no free lunch, er, brightness when it comes to boosting darks, particularly in color. Darks have limited saturation, so the expression of color at a brighter level presents a choice of saturation. We're in agreement that it becomes a subjective choice.

    Cheers,

    -Michael-

  32. #32
    Lifetime Member Jay Gould's Avatar
    Join Date
    Mar 2009
    Location
    In the whole wide world!
    Posts
    2,732
    Threads
    313
    Thank You Posts

    Default

    Attached Images Attached Images
     
    Here's my 2 cents; in the next post is the explanation. What is the identification of the bird; I would like to Google and find some images which I would have normally done before processing. I acknowledge I may have given the bird too much punch - closer to a hummingbird's brilliance.
    Last edited by Jay Gould; 02-29-2012 at 01:17 PM.
    Cheers, Jay

    My Digital Art - "Nature Interpreted" - can now be view at http://www.luvntravlnphotography.com

    "Nature Interpreted" - Photography begins with your mind and eyes, and ends with an image representing your vision and your reality of the captured scene; photography exceeds the camera sensor's limitations. Capturing and Processing landscapes and seascapes allows me to express my vision and reality of Nature.

  33. #33
    Lifetime Member Jay Gould's Avatar
    Join Date
    Mar 2009
    Location
    In the whole wide world!
    Posts
    2,732
    Threads
    313
    Thank You Posts

    Default

    Attached Images Attached Images
     
    LR: Crop only and the to CS5

    Nik Define: Defaults with brush - the mask opens in "white" reveal mode

    BKG: 100% opacity and flow touching the edge of the bird; not too careful so I could do it quickly

    Bird: 100% opacity and flow in "black" to delete any of the reduction on bird; then 50% opacity and flow in "white" on bird. When you do this you have to do it in one continuous cover and it doesn't matter if you go into the BKG. You cannot go back and do some spot on the bird a second time unless you reduce the opacity significantly because there is a coubling effect - 50% on top of 50%.

    Image: roated 90 deg CW so that the bird's wings were horizontal and I could see more of the bird to work on it; I wasn't doing anything to the BKG!

    Nik Capture Sharpener: Defaults with brush applied to bird only at 100%

    Nik CEP Detail Extractor: Brushed at 100%

    If you are not aware of Tony Kuyper's Amazing Actions, let me introduce you!! http://goodlight.us/index.html In short, these are tone based adjustment actions with only work on very limited portions of the image/histogram. Read more to learn more! For $30 it is an incredible bargain. BTW, works wonders in B&W too. Each action allows you to adjust a particular tone in either layers or curves; I like to work in curves a I just learning them.

    TK Shadow Darks: I used this to tweak the dark feathers; open them up just a bit. I put anchors along the curve line so that I was only tweaking the greyed in area of the histogram.

    TK Basic Mid Tones: Applied a shallow S-Curve

    TK Saturation Painting: Applied at 10% or a couple of passes on mid and light tones of bird.

    TK Burn/Dodge: Enlarged to 700% and dodged the white in the tail catch light at 100%

    Pixel Genius Creative Sharpener on the bird only at 50% 2x.

    Save For Web in CS5
    Last edited by Jay Gould; 02-29-2012 at 01:23 PM.
    Cheers, Jay

    My Digital Art - "Nature Interpreted" - can now be view at http://www.luvntravlnphotography.com

    "Nature Interpreted" - Photography begins with your mind and eyes, and ends with an image representing your vision and your reality of the captured scene; photography exceeds the camera sensor's limitations. Capturing and Processing landscapes and seascapes allows me to express my vision and reality of Nature.

  34. #34
    BPN Member Jerry van Dijk's Avatar
    Join Date
    Sep 2008
    Location
    The Netherlands
    Posts
    1,686
    Threads
    162
    Thank You Posts

    Default

    Attached Images Attached Images
     
    Quote Originally Posted by Michael Gerald-Yamasaki View Post
    Personally, I would like to see more comments about the various efforts here.
    Jerry - A little dark for my taste. Why the reverse-s curve (flattens midtones, not enough whites to matter)?
    Hi Michael, a little late to return to the party.....
    Good point about raising comments about the work and nice that it has been taken up!

    I like Mital's own version, although I would increase the black point a little. I'm impressed by the amount of detail Roger was able to get from the image, but agree with Elliotte that the image isn't really suitable for a close up portrait. If Roger's version had been posted as a wider crop, it would have been my favorite. I like Jay's magic on those colors, although I doubt that this species is actually that colorful. I prefer your OP over your RP, which looks much "grungier" and a little washed out.

    Re. your comments on my version: I deliberately kept it a little dark, because I didn't like the washed out BG in lighter versions (see some of the other posts here). I did a curves adjustment to bring back the details in the darker parts of the wings after increasing the black point. The curves adjustment gave me more control on what was going on compared to using the fill light slider. I locked the curve in the middle to not affect the midtones. I found that bringing down the top end of the curve resulted in a more even BG, which I liked. End result: reversed S-curve. So I adjusted the image to my liking using the curve and ended up with a reversed S-curve rather than applying a reversed S-curve as a deliberate strategy. Here's the curve:
    Last edited by Jerry van Dijk; 02-29-2012 at 02:32 PM.

  35. #35
    Lifetime Member Michael Gerald-Yamasaki's Avatar
    Join Date
    Jan 2010
    Location
    Santa Cruz, CA USA
    Posts
    1,429
    Threads
    213
    Thank You Posts

    Default

    Quote Originally Posted by Jerry van Dijk View Post
    Re. your comments on my version: I deliberately kept it a little dark, because I didn't like the washed out BG in lighter versions (see some of the other posts here). I did a curves adjustment to bring back the details in the darker parts of the wings after increasing the black point. The curves adjustment gave me more control on what was going on compared to using the fill light slider. I locked the curve in the middle to not affect the midtones. I found that bringing down the top end of the curve resulted in a more even BG, which I liked. End result: reversed S-curve. So I adjusted the image to my liking using the curve and ended up with a reversed S-curve rather than applying a reversed S-curve as a deliberate strategy. Here's the curve:
    Jerry, Thanks for posting the curve (usually reverse S implies flattened midtones but you held them steady (parallel to diagonal)), though to a certain extent it looks like you reversed the +.25 exposure adjustment with the curve (lowered with respect to diagonal) for the midtones not that that is necessarily a bad thing. BTW, your methodology is more or less how I use curves.

    Cheers,

    -Michael-

  36. #36
    BPN Member Jerry van Dijk's Avatar
    Join Date
    Sep 2008
    Location
    The Netherlands
    Posts
    1,686
    Threads
    162
    Thank You Posts

    Default

    Lowering the whole curve below the diagonal again isn't a deliberate strategy, just the result of my sloppyness in a somewhat hastened workflow.....

  37. #37
    Forum Participant Mark Itol's Avatar
    Join Date
    Oct 2009
    Location
    Philippines
    Posts
    288
    Threads
    23
    Thank You Posts

    Default

    Quote Originally Posted by Roger Clark View Post
    Hi Tom,

    The way to find clipped areas is to use the levels tool. Hold down the alt key (on windows) and move either the right or left sliders. The image will go black but as you move the slider away from the limit, if any data are clipped it will show as a color. The color and intensity depends on how many channels and how much it is clipped.

    Roger
    Hi Roger. In reference to what you mentioned re no tools in ACR to check for clipping, the above method also works for the exposure and blacks sliders in ACR. Or do you mean something different?

  38. #38
    BPN Member Roger Clark's Avatar
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,956
    Threads
    257
    Thank You Posts

    Default

    Quote Originally Posted by Mark Itol View Post
    Hi Roger. In reference to what you mentioned re no tools in ACR to check for clipping, the above method also works for the exposure and blacks sliders in ACR. Or do you mean something different?
    Mark,
    Way cool. I did not know that. That helps a lot.
    Thanks,
    Roger

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
Web Analytics