| Author | Message | ||
     Lars Tore Gustavsen New member Username: Lt_gustavsen Post Number: 34 Registered: 12-2004 |
I was very disappointed last year when I saw the result from this shoot. The sky was to blown out and it was impossible (for me) to recover it with a nice result. The I discovered this thread. https://sourceforge.net/forum/forum.php?thread_id=1430146&forum_id=434060 This is the jpg from camera.  And here is the raw file possessed in a 16bit workflow.  My workflow here is: -ufraw: for raw prosessing. -imagemagick: for the saturation blending -picture window: with the shadow and midtone part from Den's workflow . The imagemagick step is of course a feature request for pw. | ||
| Lars Tore Gustavsen New member Username: Lt_gustavsen Post Number: 35 Registered: 12-2004 |
Arg. I forgot to link to this page http://people.zoy.org/~cyril/ufraw_highlight_recovery/ I followed the ufraw part, but use imagemagik for the blending. Like this: composite -compose Saturate clipped.tif unclipped.tif final.tiff Regards Lars Tore | ||
| den Unregistered guest |
Lars Tore... This is a good example of a highlight recovery workflow utilizing the RAW file information... thank you for taking the time to describe it and link the references.... According to your 'notes' web page the jpg version was derived from a CRW file... I would like to satisfy my curiosity... if you would provide permission and a link to a large 'original' jpg file, I would like to see how much highlight recovery could be made... ...Perferably the jpg file should be minimally processed by the RAW converter so that Canon's in-camera normal/nominal settings for contrast=0, saturation=0, sharpness=0, and tone=0, and also WB remain essentially unchanged when converted to a 24bit color jpg file... I just would like to see what the '3Tone Range Adjustment Procedures' could do from the jpg file. | ||
| Lars Tore Gustavsen New member Username: Lt_gustavsen Post Number: 36 Registered: 12-2004 |
I shoot with a canon 10D. The 10D don't save a seperate jpg and cr2 file to the card, but embedded a jpg inside the crw. It's possible to change the size of the embeded jpg with a custom fuction. I have redused the size, and only use it somtimes as a reference. Here is the link to the jpg extracted from the crw. http://www.jordskifte.no/bilder/crw_2005_06_06_1548.thumb.jpg And here is the orginal crw http://www.jordskifte.no/bilder/crw_2005_06_06_1548.crw It's nothing spesial about this image, so you and everyone else are free to use it. I think I have to delete the raw version next week. | ||
| den Unregistered guest |
Lars Tore... Looks like you have posted a very creditable combination of software and workflow that will recover the highlights from a single image file... My efforts using the crw file converted to 16-bit tiff, PWP, and '3Tone Range' adjustments are illustrated here [3 images - 480KB total]: http://www.ncplus.net/~birchbay/ltg/ To produce an acceptable image, I did a 'sky' substitution... I noticed that the camera's AE was set to 'center weighted average'... perhaps a bit more of an 'expose to the right' was needed but then that becomes problematic as well as the left heavy histogram becomes even more so with resulting added shadow noise.... it will be exciting when the next generation of camera sensors become consumer available with the 'hoped for' extended dynamic range and low noise. Thank you for sharing the image.... | ||
| Kevin Elliott New member Username: Kevgermany Post Number: 294 Registered: 09-2004 |
Den, liked your version - apart from the clouds it's better imho than the one posted by Lars. However I couldn't understand the comment about the sky. Did the raw conversion blow the sky? Lars' version has a good sky, without the cheat. Can't process the raw file, so can't have a go myself. | ||
| den Unregistered guest |
Kev... to my knowledge the RAW conversion did not alter the in-camera file parameters other than convert to 16-bit tiff. It appeared to have substantially blown highlights as viewed with the Canon FileViewUtility... It would be interesting to see if one of the other RAW converters that has imaging processing features could do more before the conversion to tiff, ie, simulate exposure compensation... The Canon utility I used is very basic and left the in-camera image parameters 'as found'. The only advantage here was having access to a 16-bit tiff starting file rather than an 8-bit jpg file as I had originally requested... As to the other image tone range improvements, it may be that LarsTore's main forcus were highlights... | ||
| Lars Tore Gustavsen New member Username: Lt_gustavsen Post Number: 37 Registered: 12-2004 |
Thanks Den, it's always inspiring to see what you can do with an image. You are right, my main focus was on the highlights I guess I was too excited over the sky. I really like what you have done with the light on the trees under the bridge, I will try to recreate it soon. I guess not everybody here have access to ufraw and imagemagic (they are both available for windows). Therefor I have uploaded the file I get out of imagemagick. The file is downsized to 1/3. This is the jpg file: http://www.jordskifte.no/bilder/sogndalstrand-small.jpg This is a 16bit tif file (4.1 mb) http://www.jordskifte.no/bilder/sogndalstrand-small.tif If someone get better sky's out of the raw converter, please stand up | ||
| Steven Taylor New member Username: Staylor2 Post Number: 28 Registered: 09-2004 |
This is my try using Bibble for the raw conversion and then a color corection on just the sky using a mask in PWP and then a slight brightness curve corection on the whole picture.  Looking at the jpeg here, it looks a little more purple in the upper sky than it did in PWP before I saved it. Steve Taylor | ||
| den Unregistered guest |
Steve... thank you for participating... with the message board image posting requirements [<30KB file size], one's illustrations do not always show properly... Looks like there are the beginnings of cloud texture recovery that is not showing in the 'crw file->16-bit tiff->PWP flow I attempted but not to the same degree as 'ufraw'->'imagemagick' flow... | ||
| Kevin Elliott New member Username: Kevgermany Post Number: 295 Registered: 09-2004 |
Den, thanks for the explanation. Had a look. The ufraw is no good for me. Too much messing with open source to install. Downloaded the .tiff from lars. agree with Den, the sky's posterised, especially the clouds. Tried blurring the clouds through a cloud mask to reduce the posterisation - this helped (gaussian, radius 11), but using a sky mask through HSL - step one of the 3 tone - (yes, I've been learning Den) still tended to bring the posterisation back. However there must be something in the orignal for the imagemagick output to have so much more than the canon raw conversion. Wonder what it is? | ||
| Craig Kirkwood Unregistered guest |
This thread is amazing! Who would have thought you could do this. In response to Steve Taylor's comment about the color shift: Check out http://www.dl-c.com/discus/messages/2/2208.html . It appears that there is a slight color shift problem in the Picture Window JPEG conversion routines, which may have caused the problem. If color accuracy is critical, I convert JPEGs to TIFFs before bringing them into Picture Window. | ||
| den Unregistered guest |
Craig... have to disagree alittle with your conclusion re Steve's color shift... the original file which I worked from was 3072x2048 pixels, 37.7MB, 16-bit tiff.. which if re-sized to 386x257 pixels, <30KB, and converted to 8-bit jpg, there has got to be some changes... for example my 'substitue' sky image looks like this when this is done:  ...also, I believe that the jpeg color shifts start occuring with repeated save/open/close iterations... not with the first opening of a jpg file in PWP as it will open as an 8-bit tiff working file until you do a 'file save' at which time you can save as a jpg or tif... | ||
| den Unregistered guest |
oops.... that should be 'save/close/open' iterations rather than 'save/open/close' as indicated above. Regardless of the program, jpg save/close/open iterations should be minimized as changes are made to a greater or lesser degree by all programs with this feature, not just PWP. | ||
| Kevin Elliott New member Username: Kevgermany Post Number: 296 Registered: 09-2004 |
Den, colour shift depends on the file and colours. Red's particularly bad. I have problems on the first jpeg save, like Steve. There are quite a few threads in the BB documenting this. I have retained elements II on my machine solely for jpeg & web conversions. | ||
| Bob Walker New member Username: Rwalkernm Post Number: 14 Registered: 06-2004 |
Lars -- I processed the image through RawShooter Premium, and can recover some cloud detail, almost comparable to your example. Basically, I developed the image once for the sky, and a second time for everything else, wrote jpegs, and 3-tone processed them in PWP. A final result is attached (I hope)  I will post a more complete workflow on pbase, if anyone is interested. | ||
| den Unregistered guest |
Getting a bit OT!.... for my own needed understanding/clarification... Kev... don't disagree that some changes occur to a greater or lesser extent on the first jpg 'save/close' in PWP or any other program for that matter, but don't you think that the method used to significantly down-size the dimensions and amount of compression selected to decrease file size have the greater impact on change... regardless of the program used? Also, do I remember correctly that you can open a jpg in PWP with no loss as it opens as an 8-bit tiff working file?... so there is no need to convert to tiff in another program before opening in PWP. | ||
| den Unregistered guest |
Bob Walker... nicely done. I would be interested in seeing your workflow details... thank you in advance. | ||
| Lars Tore Gustavsen New member Username: Lt_gustavsen Post Number: 38 Registered: 12-2004 |
Here is my second attempt. (very hard jpg compression)  Here is a bigger version: http://www.mulebakken.net/div/sogndalstrand-3d-medium.jpg I struggled for a long time, until I realized that I had to expand the shadowmask tone range. Kevin thanks for the blur tips on the sky. I think I have blurred them to much, but it's to late in the evening now. Maybe, just blur the grey area will be fine. I experimented lot with the filter additive method, but I had to give up joining them together again. There are possibilities with that filter. | ||
| Bob Walker New member Username: Rwalkernm Post Number: 15 Registered: 06-2004 |
Den -- -workflow is mostly "yours". I suppose that it is possible to recover blown sky highlights using the double-develop-RAW strategy, but I do not do that often, because I don't often have images worth the effort (<smile>). There's a gallery on pbase: http://www.pbase.com/rwalkernm/pwp06a The captions on each image describe what I was trying to do at each stage of the workflow. There may be more detail there than anyone really cares about, but just in case. If you see any holes in my overall strategy, let me know. | ||
| den Unregistered guest |
LarsTore.... Another Possibility: In studying the capabilities of the Canon FileViewerUtility program more closely, it was found to have a feature to change the original Canon10D crw file exposure compensation setting. A -2EV exposure equivalent 16-bit tiff image file was created from the RAW file data and then brightness, contrast, and saturation of the highlights were adjusted in PWP with the Color Curves transform in the HSL color space. The 'highlights' enhanced image was blended with the original cameral download 16-bit tiff file in PWP using the Composite-Blend transform and a blending mask. The resulting image from this blending was then adjusted for brightness, contrast, and saturation in the shadow [0-30%] and mid-tone [30-70%] tone ranges using ‘3Tone Range’ enhancements. This enhanced image version with the original sky areas 'recovered' is shown below.  A larger version [the third image] has been added to the web page here: http://www.ncplus.net/~birchbay/ltg/ | ||
| den Unregistered guest |
Bob Walker... very well documented... I think you did a better job of masking the left side tree branches hanging over the bridge then I did... tone range selections were similar... thank you for taking the time... LarsTore... If you have Canon's FileViewerUtility, you can generate the -2EV exposure compensated file by opening the original crw file and changing EC=-2, Contrast=High, Saturation=High, and then save as a 16-bit tiff file... this is the same as if you had changed your camera settings after taking the first shot and taken another one with the new in-camera settings above.... the two exposure are then processed in PWP as described in the preceeding post.... good cloud detail is recovered which does not show all that well in the low resolution illustrations.... This has been a good exercise... Thank you. | ||
| Kevin Elliott New member Username: Kevgermany Post Number: 297 Registered: 09-2004 |
Den, Haven't done it for a while as after getting hit by the problem a few times I changes my workflow. General colour shift doesn't seem to come from resizing (but I haven't done a specific test since). There's very little or no change on converting from 16 to 8 bit(as in I didn't notice it). I noticed it after downsizing converting to sRGB, then saving a file as a jpeg, then opening it in Internet Explorer. Pulling the jpeg back into PWP has the colour shifts, while the orignal sRGB downsized tiff doesn't. I'm not sure what happens on opening the jpeg, but certainly it comes in as an 8bit editable. I try not to open/edit jpegs. Where I do get a jpeg, I open it and then save it as a tiff to avoid these and other problems. Perhaps it's overkill, don't know. | ||
| Lars Tore Gustavsen New member Username: Lt_gustavsen Post Number: 39 Registered: 12-2004 |
Bob, I have read your workflow, and I special appreciate your blending technic. You open my eyes for the overlap function. Thank you for documenting you workflow. But I wonder why you are working with jpg files, and do you do all your sharpening in the raw converter? Den, I don't have a functional FileViewerUtility, I am on linux and my main converter are bibblepro. My intention with this thread was not to start a converter war, but more point out the advantage of shooting raw. Den, what you write about -2EV and blending it together with a normal exposure from the same raw file have been discussed on this forum before. If I remember correctly Mr Sachs have told us that there is noting there that can not be done with a curve. But I think the blending tecnic can be an easy way. | ||
| Bob Walker New member Username: Rwalkernm Post Number: 16 Registered: 06-2004 |
Lars asks, "But I wonder why you are working with jpg files, and do you do all your sharpening in the raw converter?" Lars, I was just being lazy allowing RSP to do all the sharpening. The intent of the workflow was simply to focus on the tonality of the image. At ISO100, your image did not have any appreciable digital noise, so sharpening first (RSP settings even at "0" do a fair amount of sharpening) was just a easy way to avoid it later. The normal approach would to be to set RSP sliders all the way to the left for minimal sharpening, and to sharpen using PWP, toward the end of the overall workflow. Once you are in 8-bit color space (jpeg), of course you cannot recover highlight details in the way that is possible using 16 bit colors. So that is why I went with the dual RSP developing. I agree (with Jonathan) that curves in 16-bit space should be able to accomplish the same thing, if perhaps more difficult to isolate what you are doing at each stage of the processing. I used the jpeg files because others did not have access to RAW converters, and I chose RSP (over Capture One Pro) because RSP is more generally accessible for anyone who does not already have one. The RAW converter is the only stage in the process where we can access the full 10- or 12-bit data from the sensor, and once we have converted to jpeg, we are in 8-bit space. I am not worried about lack of color faithfulness, the was only one cycle of transforming in and out of jpeg color space (and PWP jpeg, while different from many other photo editors, is not significantly worse than others if that is all the jpeg transforming you do). | ||
| Kevin Elliott New member Username: Kevgermany Post Number: 298 Registered: 09-2004 |
Lars/Bob Den's 3 tone does things that curves doesn't and can't do. where curves bring things out, they affect the overall image and an increase in highlight detail/range means a compression of the range of the remaining image. 3 tone allows you to concentrate on a specific tonal range and thus stretch it more than you could with curves without adversely affecting other parts of the image. If you look at the image recovery that Den's achieved, then try and do the same thing with curves, you'll find it's not possible. | ||
| den Unregistered guest |
Kev... thank you, you explained it well... my last workflow was not really exposure blending but highlight tone range replacement of the 70->100% tones of the original crw file. OT...Also Kev... don't spend time re jpg color shift/quality issues... they are what they are and have been more than discussed elsewhere... LarsTores and others..... I normally do not shoot raw prefering to acquire the best upfront camera exposure and then utilizing the resulting camera download file as the highest quality/size jpg file. Had this photo been re-shot changing the camera setting exposure compensation, EC, to -1EV, the resulting jpg camera download image would be properly 'exposed to the right' [see: http://www.ncplus.net/~birchbay/tutorials/camera_exposure/01.htm ] but with a left heavy heavy histogram requiring shadow and lower mid-tone recovery of the '3Tone Range' procedures plus specific image area enhancements [See: http://www.ncplus.net/~birchbay/3tone3/shdw-midtns.htm ]. Using Canon's FileViewerUtility [part of there solutions disk programs that come with their cameras... do not mean to start a RAW converter war but it is the only converter I have experienced], I simulated the in-camera programing change to EC=-1EV using the original crw file and converted the resulting file to an 8-bit jpg file with best quality. This is normally the start image for the '3Tone Range Adjustment Procedures'. This resulted in producing the following final image suitable for a photo quality print:  With an 'exposed to the right' jpg camera download image and the '3Tone Range' procedures for PWP post-processing, there would be no need for RAW conversion or the other 'Herculean' efforts that have been necessary for other workflows... If there is any interest, I will upload to a server, a medium resolution version of this image for your inspection... | ||
| Tony Gamble New member Username: Tonygamble Post Number: 153 Registered: 04-2003 |
Den, Your spam blocker seems to be blocking me so I'll ask here... In your latest edition of the Three Tone paper you introduce a point b to identify the limit of the highlights to mask and a to identify likewise with shadows. Could you explain in the paper where the Probe should be place in the image to pick these spots please. Rgds Tony | ||
| Lars Tore Gustavsen New member Username: Lt_gustavsen Post Number: 40 Registered: 12-2004 |
Den, I normally try tho shoot to the right, but not this time. Thanks for the link to your shdw-midtns.htm. Looks like my file responds very nice to the filter transformation. Like Tony Gamble, I think some guidelines for picking a and b parameter would be a nice thing. I would also like to see for step2B (the shadow brightness curve) a 100% crop showing “halos/auras just appear” | ||
| den Unregistered guest |
Thought for sure someone would take exception to my 'jpg file is adequate stance' but be assured that this thread has re-envolved me with RAW conversion aspects as well as the added image data content of 16-bit vs 8-bit files... it is just that my 'ancient' computer system takes too long to do number crunching, so for the most part, I work with 8-bit files... especally since they would get converted to 8-bit anyway by printer and monitor drivers for print/display.... TonyG, LarsTores, and Others: Picking '3Tone Range' Transitions: 1.A: The specific tone range transition tones are image content dependent and preference driven depending upon the desired end results... this is why this process is so difficult to 'automate'. The human mind is the best to make this decision then a set of generalizing algorithms. There are some trends that I have noticed: Shadow Ranges: 0% to 20->40% and Highlight Ranges 60->80% to 100%. Another good place to start is locating the transition tone in a histogram dip or low spot. 1.B. Tone Range curve selections for brightness, contrast, and saturation are particularly image content and preference driven... again so much so that I really don't have any generalization beyond what has already been presented... except to zoom the 'Preview' and keep pushing the curve points around until something is acceptable... For both 1.A and 1.B, initially, trial and error play large roles in the learning process... with practice, one learns what works and what doesn't... practice is why I enjoy threads like this... For your consideration, I have written a tutorial using a 2Tone Range approach for LTGs 'bridge' image based upon a starting image that is -1EV EC from the original crw file. This tutorial with a work-along image is here: http://www.ncplus.net/~birchbay/tutorials/LTGs_bridge/01.htm As you work you way through the steps, do change the transition tone [from 70% to 65% or even 60%] as well as the suggested curves...watching the changes in a zoomed 'Preview' window to draw your own conclusions as to what is being affected and what is producing the most pleasing results. This is wrote with only one transition tone so there are only 2 tone ranges... just to keep things a bit simpler... concentrate on Steps 1 and 2. Step 3 is image sharpening and Step 4 is image area considerations for the silhouetted tree/limbs above and left of the bridge. While this work flow only has two tone ranges, previous work flows had shadows=0-30%; mid-tones=30->70% and highlights=70->100%. BobW in his workflow chose: shadows=0-20%, mid-tones=20->70%, and highlights=70->100% with a unique masking technique worth consideration for the tree/limbs above left of the bridge [See '80_ScreenSkyMask2' of his work flow]. Where he chose to do a color change to remove the 'light green' cast above the bridge rails, I chose to lower the saturation. Have some fun.... | ||
| Kevin Elliott New member Username: Kevgermany Post Number: 302 Registered: 09-2004 |
When picking the highlight boundary, I try to avoid any value that's part of the sky. In general I tend to set the boundary slightly below the darkest point of the sky. As Den says, there are often a lot of clues in the histogram. For shadows in an image where I haeve difficutly identifying a transition point I use the grey/brightness tool, play till the shadows look right - ignoring the effect on the rest of the image, then estimate the transition point from there. This boundary is the lightest point that you want to lighten and bring into the mid tones. | ||
| Tony Gamble New member Username: Tonygamble Post Number: 154 Registered: 04-2003 |
Den, I'm sorry but you did not understand my question. In your Three Tone paper you use the sentence:- "Use the Probe to determine the needed range start tone, b." Later you say:- "Use the Probe to determine the needed range end tone, a." I repeat what I said in my post. "Could you explain in the paper where the Probe should be place in the image to pick these spots please." I think that is also what Lars was asking. Thanks in anticipation. Tony | ||
| Dieter Mayr New member Username: Dieter_mayr Post Number: 241 Registered: 11-2004 |
Since no one mentioned it before ( or i just overreaded it ): The eyedropper probe is very useful to find the points in histogram, and Shift-Click sets a control point in historam. Kevin, thanks for the tip with Brightness, works very well. Dieter | ||
| den Unregistered guest |
Ahhhh... the light goes on! As Dieter alludes, click on the Mask Tool - Brightness Curve - Probe option... then do a mouse 'left-click-hold', moving the mouse around on the image... a vertical bar moves around in the Mask Tool - Brightness Curve histogram indicating tone value... a mouse 'shift-left-click' will add a control point to the curve... and incidently, a mouse 'control-left-click' will remove it. I use this initially to analyze an image's histogram and I take it for granted. This 'probing' is necessary to surgically extract a desired tone range... in this case, by mouse hold-left-click, you can see that if you go much below 70% tone that you are no longer in the bridge image's sky area hence this becomes the starting 'b' value... The image does not have a well defined 'a' value, hence I was able to simply to a 2Tone Range solution rather than a 3Tone Range solution... or the shadow range becomes a preference selection as KevinE describes. | ||
| den Unregistered guest |
This same probe technique is also used with the Color Curves or Brightness Curve transforms in selecting, adding, and/or removing curve control points.... for example see the screen shot for Step 4, Workflow1 here: http://www.ncplus.net/~birchbay/3tone3/shdw-midtns.htm | ||
| Tony Gamble New member Username: Tonygamble Post Number: 155 Registered: 04-2003 |
Thanks Den, The interactive use of the Probe will help us understand how those masks evolve much more clearly than simply working from figures such as 70, 80 etc. And I always find remembering the techniques comes far more easily once I see the logic instead of trying to learn them 'parrot fashion'. Tony | ||
| Lars Tore Gustavsen New member Username: Lt_gustavsen Post Number: 41 Registered: 12-2004 |
Thanks Den for a very nice writeup. I have followed your workflow with your image. The resulting image is very nice. The colors looks very realistic. (I printed it to evaluate it). I just felt that I starting understanding the mask brigthness tool, and then you introduce this hard color curves. I guess I have to study HSV and HSL colorspaces more. Learning is always fun. Your step 4 are is really great. Everyone else: After I have saved a lot of curves with starting point, the complex masking are faster. I have saved curves like this: shadow.crv and shadow-subtract.crv. Then in the mask tool, load the curve and slide the controlpoints away. | ||
| den Unregistered guest |
Lars Tore/Others... Addendum 1 has been added to the 'bridge image' tutorial illustrating a '3Tone Range' Step 2B Brightness Curve added to the 0->20% shadow tone range.... http://www.ncplus.net/~birchbay/tutorials/LTGs_bridge/01.htm#Add1 Hope this helps with understanding of the range of possibilities and that there are no limits other than one's own... | ||
| Robert W. Coutant New member Username: Couman Post Number: 99 Registered: 01-2005 |
This has been a very interesting and informative thread, especially when considered along with parallel discussions that have popped up on other forums. The general conclusions seem to be that: 1. Significantly more information can be gleaned from a single RAW file than from any given derived image (jpeg or tiff) file. and, 2. Exposure bracketing can be simulated by alternative conversion of a RAW file to different image (jpeg or tiff) files. The first point is certainly valid when the derived image is in 8-bit format. In my limited experience, I’ve not seen any hard evidence for or against the equivalency of 16-bit tiff, but my intuition leans towards equivalency. On the other hand, there is strong evidence that different RAW converters can produce different results from the same starting files. The second point is acceptable as long as we emphasize “simulated”. Variable development of a RAW file can certainly help bring out detail in an underexposed region of the image, but it cannot reveal as much detail as would be present in a perfectly exposed image of that same region. If the latter were true, there would be no need to release the shutter multiple times for bracketed exposures – it could all be done by the camera firmware. Finally, there is software, other than that mentioned in this thread, that offers alternative approaches – see http://hdrsoft.com/ for a discussion and illustrations. [As Den has stated, each image is a separate case and each of us has different preferences, but it's interesting to see how others approach this kind of problem.] Bob C. | ||
| Kevin Elliott New member Username: Kevgermany Post Number: 303 Registered: 09-2004 |
Bob, don't quite agree your second point. What we're trying to achieve is to reduce a huge real life brightness range into the constraints of a screen or print. The raw converters seem to compress and thus smother detail, present in the camera. I did download and install RSE to play with Lars' image. The differential exposure preserves both end's detail and blending gives a better, more realistic image. This is similar to adjustments I routinely carry out with Den's 3 tone - or in the grey/brightness tool for less demanding images. I look on it similarly to getting cheap lab prints back from negs, with blown highligths,and blocked up shadows - then checking the neg, scanning it carefully and then bringing it back in PWP - great prints with full range of shadow and highlight details. Pre scanning & PWP, the printer would have dodged and burned at the printing stage to achieve the same effects. So to me this isn't a simulation, merely a way of getting the most out of the shot. However There's another issue here, not factored in and that's what happens when the scene brightness range is too great for the sensor. in this case bracketed exposures are necessary, probably combined with even more post processing. Lars, thanksfor posting, I've learnt a lot and I'm sure others have as well. Now, following Den's private instruction, I'm going to study Bob Walker's approach. | ||
| Robert W. Coutant New member Username: Couman Post Number: 100 Registered: 01-2005 |
Kevin, perhaps my use of "simulate" was a poor choice of words. I agree that, for this image, and probably many others, the approach is useful. However, it does not and cannot do the full job in the case where the sensor's capacity has been exceeded. Again, we are in full agreement. Bob C. | ||
| Bob Walker New member Username: Rwalkernm Post Number: 17 Registered: 06-2004 |
A post in the more-than-you-ever-wanted-to-know mode, perhaps. Thoughts on setting the control points using the 3-tone image processing in PWP. In the last couple of months I have processed a few hundred images using the 3-tone image processing workflow that Den has illustrated so well with his tutorials. Here's a few observations about what I have discovered about the process. I use the 3 tone process as a way of enhancing the apparent dynamic range of an image. Obviously it works best when the original exposure of the image is good (as opposed to the sky recovery scenario that started this thread off). The apparent expanded dynamic range comes mostly through the midtone mask, and I think of the highlight and shadow masks as damage control for the extreme tonalities of the image. For that reason, I like to have the midtone mask cover as much of the image as possible, and restrain the highlight and shadow masks to cover a smaller part of the image. General principles: Each of the tone masks is the combination of an "on mask" and an "off mask". The gradient for the on mask is the steeper of the two, because you want to define more sharply what basic tone range you are working with. After blurring the on mask, you build a broad halo around its lower limit using the off mask, and so the gradient for the off mask is less steep. The highlight mask: I set the control point so that the highlights that are masked cover the areas of the image that (a) need no more contrast enhancement, and (b) can even use a little detail recovery by bringing their tones down a bit. The only way to evaluate where to put the control point is to use the probe tool, or to look at the mask and experiment. Using the HSL color space to bring the highlights down has the additional effect of adding a little saturation to the highlights, so this saturation enhancement is an added bonus. The shadow mask: The reasoning here is pretty much the reverse of the highlight mask. I try to identify dark portions of the image that will benefit from lightening up a bit, and the HSV color space in this step also keeps the saturation up in the lightening process. The midtone mask: To a zeroth order approximation, the midtone mask covers what was was not included in the highlight and shadow masks. This is only a very rough approximation, and it is not at all true in detail. In fact, the overlap between the midtone mask and the extreme tone masks needs to be reduced as much as possible to avoid funny looking transitions between these tone ranges ("auras" and "halos"). It is keeping this overlap down that makes it nice if there's a local minimum in the image histogram (the mask histogram shows the V value of the HSV space), because then you minimize the area of the image that you are adjusting with multiple masks (yes, many pixels in the image get adjusted by all three masks, as an effect of the blurring step in making each mask). I make all the masks first, while I can keep in my head where I set the control points. For 80% of the images (6MP, Canon 10D), I set the control points to: On Mask Off Mask Blur Radius Highlight (70,0)to(80,100) (50,100)to(70,0) 40-60 Shadow (20,100)to(30,0) (30,0)to(50,100) 40-60 MidTone On: (30,0),(40,100),(60,100),(70,0) (50-100) MidTone Off: (10,100),(30,0),(70,0),(90,100) The adjustment of the blur radius follows Den's tutorials: look at how big the blobs in the mask are, try not to wash out entire areas of importance by blurring them to grayness. After the masks are made, I then do the three brightness control adjustments. I doubt that the order is crucial, but I always do (in the order) highlights in HSL, shadows in HSV, and midtones in RGB (although I can rarely tell any difference between RGB and HSV). The highlight step brings the high tones down toward the middle of the histogram, and the shadow step pushed the low values toward the middle. The midtone step pushes the middle of the histogram to the outsides. More about this later. A final note: most of the images I did this with are pictures of kids playing soccer. Pictures tend to be in the bright sunlight (you hope), which makes for very contrasty images (and the 10D dynamic range is OK, but not remarkable). So the highlights (bright white ball, sky, sun reflecting off player's skin) are improved by bringing tones down. Shadows (where the sun casts a shadow on the field) are helped so that you can see some of the grass, and the midtone corrections remove a general haze from the image. The thousands of images on my pbase site (www.pbase.com/rwalkernm) have NOT been 3-toned, because it would take me years to do that. But when parents ask for copies, I clean them up and it works very nicely indeed. Another final note: This is a bonus for both of you who have read this far. There is a fourth tone mask you might find interesting to look at. Open up your original image, and apply in succession using the combine masks tool in add mode, a copy of all three of your masks. Invert that. What's left is a mask that roughly shows the (spatial) regions of the image that were NOT affected by ANY of your three tone masks (or were only partially affected). It is interesting to think about. This mask shows, for the most part, the regions of your image that are in the two overlap regions BETWEEN the midtone mask and the extreme masks. So maybe a way to think about this process is that we are pulling the highlights down toward the upper overlap region, pushing up the shadows toward the lower overlap region (near the zero control points), and with the midtone mask we are pushing the center of the tones towards both overlap edges. I find that curious, and the final histograms don't look that way anyway, but that's what sticks in my head. | ||
| Kevin Elliott New member Username: Kevgermany Post Number: 304 Registered: 09-2004 |
Bob, went through your pbase example last night. Very interesting. My final results were similar to yours, although I must confess that I didn't think of masking teh clouds to reduce colour saturation there - I tried a colour correctoin, unmasked onthe overall image and tried to correct the magent there, unsuccessfully. I like your approach of generating all 3 masks first - it lessens the effect that I've sometimes seen following Den's method, that teh mid tone adjustment can push back the shadow/highlight adjustments where the masks overlap. One thing I have found is that shadow recovery can lead to oversaturation, this is (to my taste) evident in some of the versions of Lars' shot posted here. Sometimes it's better to do the midtones in HSV to get all the saturation levels the same during the 3 tone, then adjust all together for overall effect after 3 toning is complete. | ||
| den Unregistered guest |
BobW... I appreciate your knowledgeable treatise/analysis and feedback regarding your use of the 3Tone approach as initially presented... Truthfully, I personally have not gone much beyond what I have already presented... just become more adept. One of the my friends that I correspond with off the message board also found that doing the tone range that includes the most image area first was best for him.... from your description the sequence would be: mid-tones, then the highlights and shadows.... My original thinking upon building the tone range masks upon the successive working image was to maximize retention of their respective tone range enhancements and I have had to go back and re-do a tone range step because of an unacceptable transition zone [usually changing the amount of mask blur]... I will give your suggestion of developing all masks upon the base image a trial. Have you tried using the '4thMask' to correct the transition zones? Whenever I have combined masks with significant gradients, the combined gradients seem to form nexus/nodes rather another smooth gradient... probably has something to do with PWP's inability to have a 'soft edged brush erase' capability... Again, I will start taking a look at the '4thMask' and it may very well be un-necessary to modify transition zones where the the masks are based on the original image. The '4thMask' is intriging as it may be useful where the transition zone occurs in a troublesome image area... the silhouetted tree limbs/leaves of the tree above-left of the bridge for example... Lars Tore.... Thank you for exposing your image to such close scrutiny/dissection... going from an 'open source' highlight recovery to overall enhancement... this kind of discussion is needed to further one's understanding of the nature/manipulation of image data... so again thank you. Last but not all least... Thank you to all participants... | ||
| Tony Gamble New member Username: Tonygamble Post Number: 156 Registered: 04-2003 |
Bob, You use "On Mask Off Mask Blur Radius" as per the table below. Does that mean that you don't feather, as per Den. Tony | ||
| Bob Walker New member Username: Rwalkernm Post Number: 18 Registered: 06-2004 |
Kevin, yes I do make all 3 masks first. To make the high/low masks after the midtone mask is applied seems wrong -- because the actual mask will depend on how much brightness transform correction you applied. But I admit I have not tried that -- and I see that Den does it successively. I guess I need to try that out, maybe Den's way is more resistant to developing auras and halos. Den, maybe the reason I do the largest midtone mask last is another consequence of making all 3 masks first. Another idea for thought. Of course, making all three masks first could just be laziness, trying to avoid hopping around between windows. I don't think there's anything you can do with the transition tone mask, because it should contain a very broad histogram of tones. But if you do end up with regions with auras and halos, I bet they coincide with places where the transition mask is bright, and where values corrected from an extreme mask have been transformed to fall on the wrong side of the corrections applied by the midtone mask. Places that end up looking funny for me are often in parts of the image that are out of focus (depth of field), where the bokeh from what would have been specular reflections does not fall off "right". Tony, Sorry, I was incomplete -- off course I do the feathering, that step seems like polishing the mask. I do the -2 first, then the +2, but I like the suggestion of filling in pinholes with the +2 first. I may play with that some, but generally there are not many pinholes because of the blurring of the "on mask". The pinholes are "restored" by application of the off mask. I am writing a routine to compute the mask-weighted histogram of an image. I think that will be instructive -- because then we can see what tones are being selected by each of the masks. Mask weighting just means this (to me): Each pixel has an V value (in HSV). The normal V histogram (which is what you see in the mask histogram dialog) counts (as the height of the y axis) the relative number of pixels with the tone value of the x axis. Each pixel contributes a value of 1 to the y axis. To make a mask-weighted histogram, just make the value a pixel contributes equal to the mask brightness (0 for off and 1 for on, and in between for partially selected pixels). Maybe Jonathan could make that an option -- wasn't there a request for a histogram of the selected pixels in a mask? Thanks all for all the ideas. Bob | ||
| den Unregistered guest |
BobW’s ‘4thMask’ suggestion [last paragraph here:http://www.dl-c.com/cgi-bin/discus/show.cgi?tpc=2&post=12434#POST12434 ] is powerfully elegant and should/will become part the ‘3Tone Range Adjustment Procedures’ for me if there is difficulty in the transition zone. For example, in LarsTore’s ‘bridge’ image, the silhouetted tree limbs above-left of the bridge and bridge railings are problematic… the limbs become over-saturated and both the limbs and railing/lampposts tend to become easily over-sharpened [strong sharpening auras]…. I approached this with a hand-painted-blurred mask in the tutorial Step 3A [sharpen non-limb/railing image areas] that was later modified in Step 4A [decreased the transition contrast and saturation in the tree]… but look at the '4thMask’ [Step 1A mask + Step 2A mask; INVERT, Blur=5, expand to Full Range in Levels and Color transform] in comparison to the hand-painted mask:  Much more elegant…. Thank you Bob Walker. Caution: Pin-holes in a tone range mask may be there for a purpose. Make sure of the purpose of the mask…. Are you making adjustments for tone in a specific tone range or making adjustments to an image area?… There is a difference especially since the tone range masking technique can be used to easily select difficult/complex image areas like ‘silhouetted tree limbs’. The few trials I have done regarding the issue of making tone range masks based upon the base image or successive working images is inconclusive [at least for me] as is the order in which the tone range changes are made…. These may be personal preference choices. …..but it does seem that as the shadow [0->30%] and highlight [70->100%] tone range band increases beyond 30% width that they take on a predominate role in the tone range drama of image content and should be done first before the mid-tone range. Also I still feel intuitively that tone range masks should be formed upon successive working images to fully realize the tone changes being made in the image….. for example, I could not re-do the ‘soaring on the bay’ image [Today’s Bonus Photo II, http://www.ncplus.net/~birchbay/IMAGEsbyDEN.htm ] with the same preference results if the tone range masks were made from the base image. For this image the shadow and highlight bands were 35% and 40% wide respectively. Additionally, the contrast in the shadow and highlight tone ranges were more important than the mid-tone range. …Ignore the ‘seagull’, it was composited in… They were soaring around that afternoon on the high COLD wind and I could not get a single photo where a gull and the background were in sharp focus at the same time with the 300mm lens, so I made a composite… in other words, I gave up waiting and went back into the house for a 'hot' cup of coffee! | ||
| den Unregistered guest |
Addendum 2 has been added to the tutorial illustrating final image version changes when Addendums 1 and 2 have been employed. The differences are subtle but present in these medium resolution images. http://www.ncplus.net/~birchbay/tutorials/LTGs_bridge/01.htm#Add2 | ||
| Bob Walker New member Username: Rwalkernm Post Number: 19 Registered: 06-2004 |
Den, Thanks for the addendum example. I guess I am in danger of learning something new again. My interpretation of the 4th tone mask, as a mask that identifies tonal ranges not affected by the first 3 (or two if 2-tone) masks, seems only to make sense if all the masks are built from the base image. Your results and others seem convincing that the better way to make the masks is successively, and so that makes it harder for me to figure out what the 4th mask is showing. But, if the 4th mask for you hits the spots that still needed attention, that's evidence enough that something interesting is going on. Your 4th tone mask is different than mine was, but we did things differently. Started with different images, applied masks with different settings and in a different order. So many things to think about. However, changing the subject slightly, it does seem that you want to apply the midtone mask last in the sequence, for images where (a) the purpose of the highlight and shadow masks is to push tone values more toward the midtone region, and (b) the purpose of the midtone mask is to increase contrast by pushing tone values to the edges. If you do the midtone mask first, and if the extreme masks are narrow tonally, then I would worry you risk posterizing the image. Maybe in real life it isn't a problem, but that's why I have always done the midtone mask last. Thanks again, Bob | ||
| den Unregistered guest |
BobW... DENs '3Tone Range' Clarifications Summary/Preferences: 1. Tone Range Enhancement Sequence for Landscape/Seacapes: Highlights in HSL color space, Shadows in HSV color space, and Mid-Tones in either HSV or RGB color space [changes in the RGB color space will generally be more vivid and similar to the coloration extremes that PhotoShop editors will provide while changes in HSV,V and HSV,S can be more controlled to reflect photograph reality]. 2. Tone Range Masks based upon successive working images starting with highlights, then shadows, and finally mid-tones. Re: the '4thMask'.... I look upon the '4thMask' as a near representation of the transition zone's tonal graduations and location when it is derived from tone range masks based upon successive working images... as such it can be modified and used as an 'area mask' to decrease auras/halos, control sharpening, or other misc touchup that I have been doing by hand with the Clone Tool and Misc Tools. While this still may be needed, it is greatly simplified as demostrated with the 'bridge' image... when I applied it, I erased the spot under the bridge and used the mask as an 'area mask' to subtlely change the transition zone contrast [decreasing halo] and to de-saturate the bright 'blue' out of the tree that got in there because of the aggressive saturation increase in the HSL color space for the sky. I have clarified the my preferred tone range work flow sequence in the tutorial ADD1 statement.... | ||
| den Unregistered guest |
Sorry if the previous post seems abrupt but felt the need to established a basis of discussion... At the moment, I do not see anything that changes this basic stance and it is currently what the '3Tone Range Adjustment Procedures' tutorial proposes... BobW's '4thMask' concept [understood or not] when derived as described in the previous post, provides the near location and approximate tonal graduations of the transition zone between two tone range enhancement image areas.... Should the enhancement of these two joining ranges result in large mis-matches of brightness, contrast, and saturation over a short gradient [inspite of the original tone range mask's ADD/SUBTRACT slopes and Blur], then this mask may be used to great effect to further blend the two ranges together... Previously, I had no good way of doing this other than to hand paint-blur a mask and then use a 'curves' transform or use the Clone and Misc Tools or... going back and re-doing the original tone range mask's blur amounts. This is what make Bob Walker's '4thMask' so elegant for me... not that it is analyzing or accurately portraying the image region but it gives me closer approximation than doing it by eye-ball and hand. Thanks again BobW! | ||
| den Unregistered guest |
As an additional enhancing technique for Lars Tore's 'bridge' image, I offer the 'dappling effect' described/illustrated here: http://www.ncplus.net/~birchbay/tutorials/LTGs_bridge/Add3/ Here is an illustrated 'roll over' of my workflow's final image version and LarTore's original 'crw' file. http://www.ncplus.net/~birchbay/tutorials/LTGs_bridge/Add3/index.htm#final This has been an interesting and educational journey.... Thank you Lars... hopefully we did not loose you... If there are questions, ask... I will be off line for several days starting Sunday morning, Mar 5 but will be here for the remainder of today, Mar 4. | ||
| penumbra New member Username: Penumbra Post Number: 5 Registered: 06-2004 |
I decided to try a PWP-only analog to the multiple raw conversion technique for recovering sky detail. I filtered the photo with itself twice with 0 exposure compensation to generate a suitably underexposed image, increased sky contrast slightly using a brightness curve, then combined the resulting sky with the rest of the scene from the original image. -tom  | ||
| Bob Walker New member Username: Rwalkernm Post Number: 24 Registered: 06-2004 |
Tom, Good job, I can also reproduce that approach from a single 48-bit RAW development. If you apply a filter transform of the image with itself, it is very close to using a parabolic tone curve in HSL space. Doing it twice makes the final effect look quartic. Alternatively, you can apply a tone curve in the HSL space, and combine it with the original image using a 3-tone mask for the sky. To my way of thinking, this reinforces (proves?) the idea that a single 48-bit RAW development captures everything you need to process the image in PWP. Bob |
