If you really want to make an image stand out, you have to spend some time doing D&B in post. D&B enhances local contrast in brightness, hue and saturation.
mshi wrote:
If you really want to make an image stand out, you have to spend some time doing D&B in post. D&B enhances local contrast in brightness, hue and saturation.
Took me a while to figure out that you meant dodge and burn. Of course, D&B only enhances local contrast if you apply it in the right places (you can also decrease local contrast if you want).
There are other simple and very effective ways to increase local contrast, such as applying unsharp mask at low strength and high radius (10+ pixels radius).
carstenw wrote:
I think I remember the thread you mean, but in the end, what I am looking for is detail in the downsized image which looks like the detail in the large image. If it looks like that and has few artifacts, it is great. I don't get this from single-step solutions, but I do get this from multi-step solutions. While I do not know that the detail I get is somehow "original", I don't see any particular reason to doubt that it is, since it is repeatable with the same process and many different images with different properties.
We could compare approaches in this thread. I made a composite image (3600x2400 px) using five diverse originals (unscaled crops from raw files) plus a ring target.
Rules:
To be downscaled to exactly 1200x800 px.
Global processing of the image as a whole.
If people feel that different panels require different approaches, they can present multiple results.
I will kick off with what I consider to be a good balance between looks and artifacts. The method is downsizing with the lanczos3 kernel (a single operation, in this case in Matlab), followed by a mild high-pass sharpen (in PSP X3):
AhamB wrote:
Took me a while to figure out that you meant dodge and burn. Of course, D&B only enhances local contrast if you apply it in the right places (you can also decrease local contrast if you want).
There are other simple and very effective ways to increase local contrast, such as applying unsharp mask at low strength and high radius (10+ pixels radius).
Yes, there are many time-saving shortcuts, such as the one you mentioned, since essentially all sharpening is about altering local contrast . However, nothing beats the quality that D&B can produce. Of course, retouching has more to do with decision-making in the first place.
Toothwalker, I see that there is a problem with the last one, but a couple of the first five look soft in your example, and sharp in James' example (maybe a little too sharp, in fact). I think it isn't necessary to be so afraid of a little false detail that you leave things soft.
Here is a step sharpening version tuned for minimum artifacts in the ring shot.
Here is probably what I would have ended up with. Note that the original is sharp already, which I am not usually dealing with. This is probably what makes a couple of the shots look very slightly over-sharpened.
With my experience post processing technique must be adjusted by what camera is used, not the lens.
One has to also consider that when making images for web presentation, theres currently a very large difference on screen resolutions, on the other end theres a people using minilaptops and iphones for web browsing and on the other end theres 30" high quality screens. Photographers usually use screens with high pixel density so most of the images are bit oversharpened for the majority of the viewers.
Sami Ruusunen wrote:
One has to also consider that when making images for web presentation, theres currently a very large difference on screen resolutions, on the other end theres a people using minilaptops and iphones for web browsing and on the other end theres 30" high quality screens. Photographers usually use screens with high pixel density so most of the images are bit oversharpened for the majority of the viewers.
Indeeed. I was a bit disappointed with the pixel size of my 24" 1920x1200 screen when I first got it (coming from a 30" 2560x1600 screen). It makes sharpening artifacts more visible.
@Carsten: To me, your version of the composite image doesn't look very sharp in the fine details, but kind of coarse and grainy. James' version looks a bit cleaner, IMO.
mshi wrote:
Yes, there are many time-saving shortcuts, such as the one you mentioned, since essentially all sharpening is about altering local contrast . However, nothing beats the quality that D&B can produce. Of course, retouching has more to do with decision-making in the first place.
The achieved quality completely depends on your PP skills, doesn't it? Personally I find it hard to use D&B without being able to see the brush strokes in the result. It happened to me that when I read that a nice image was treated with D&B, I suddenly saw the brush strokes, which annoyed me.
Using the sharpening method with large radius is more of a global adjustment for small structures (not entire subjects because that will require huge radius and will produce haloes I think).
I found NIK Viveza very useful to make subjects stand out, by the way. It has a very competent technology to select complex subjects.
Toothwalker wrote:
I made a composite image (3600x2400 px) using five diverse originals (unscaled crops from raw files) plus a ring target.
Seriously? 8-bit non-color managed übersharpened original... I really hope you are not serious, as a joke this is quite fun thou.
I'll see what can be done later tonight, if I figure out how to do gamma 1.0 processing with gamma 1.8-2.5 image (hard to eyeball what gamma you have used on your non-color managed image) without assigning some ICC to it (since then my image would look different than any other image). This nightmare will be blast from past, I have been having fully color managed workflows almost decade now...back to 90's
AhamB wrote:
Indeeed. I was a bit disappointed with the pixel size of my 24" 1920x1200 screen when I first got it (coming from a 30" 2560x1600 screen). It makes sharpening artifacts more visible.
@Carsten: To me, your version of the composite image doesn't look very sharp in the fine details, but kind of coarse and grainy. James' version looks a bit cleaner, IMO.
I presume you mean my second shot? I was trying to minimize halos and artifacts in the ring image, while doing the sharpening. Perhaps if you don't see halos in James' image you are using a monitor with a different pixel pitch than I am? If you are still on the 24" screen, your pixels are a lot larger than mine.
Samuli Vahonen wrote:
Seriously? 8-bit non-color managed übersharpened original... I really hope you are not serious, as a joke this is quite fun thou.
I'll see what can be done later tonight, if I figure out how to do gamma 1.0 processing with gamma 1.8-2.5 image (hard to eyeball what gamma you have used on your non-color managed image) without assigning some ICC to it (since then my image would look different than any other image). This nightmare will be blast from past, I have been having fully color managed workflows almost decade now...back to 90's
Samuli
It's going to take a while, but wait and I will see what I can do for you.
Samuli Vahonen wrote:
Seriously? 8-bit non-color managed übersharpened original... I really hope you are not serious, as a joke this is quite fun thou.
I'll see what can be done later tonight, if I figure out how to do gamma 1.0 processing with gamma 1.8-2.5 image (hard to eyeball what gamma you have used on your non-color managed image) without assigning some ICC to it (since then my image would look different than any other image). This nightmare will be blast from past, I have been having fully color managed workflows almost decade now...back to 90's
It is not a joke, just a comparison of methods using the same input image - not necessarily ideally prepared for any particular method.
I saved four of the images with Lightroom, which I only recently started to use. I did not even realize that it applies sharpening by default. In the camera I have set zero sharpening, which DPP automatically adopts but LR apparently not. The basket was taken from an existing TIFF file that was indeed sharpened for the large format.
Concerning the 8 bits, that is the result from starting with an 8-bit ring chart and the fact that it does not matter for my processing, but I will happily provide you with 16 bits.
Here is a new version. The originals were saved as 16-bit TIFF files from LR4 with all sharpening options set to zero.
It was a pain to crop the images to precisely the same pixels as yesterday, and an even greater pain to figure out the precise formula for the ring chart (which I took from this site) but now even the ring chart is a 16-bit original. Of course, after all this pain I expect you to demonstrate that you need 16 bits.
I can't help you with color management, because I don't do color management.
Edit. Hmm. I see now that saving 16-bits with Matlab gives a 16-bit TIFF file with only 8 real bits.
I removed the file and will try to fix it later.
Edit2. It should be OK now.
Edit3. The file is too big to keep on the server forever.
Samuli Vahonen wrote:
Seriously? 8-bit non-color managed übersharpened original... I really hope you are not serious, as a joke this is quite fun thou.
I'll see what can be done later tonight, if I figure out how to do gamma 1.0 processing with gamma 1.8-2.5 image (hard to eyeball what gamma you have used on your non-color managed image) without assigning some ICC to it (since then my image would look different than any other image). This nightmare will be blast from past, I have been having fully color managed workflows almost decade now...back to 90's
Samuli
Quite useless unless you compensate for the exposure curve also, in all the natural images. The synthesized image (the zone plate) is non-treated.
The need for gamma 1 in large scale ratios is not to be underestimated. Scaling gamma-treated images will in all cases over-emphasize dark-on-bright detail in the finished resampled image - making it look slightly bloated. The same goes for larger amounts of sharpening, that when done in gamma two-point-something will oversharpen dark detail and hardly touch bright detail. This adds to the "sharpened" look you often see.
I assumed simpified sRGB gamma and went from there.
One very typical effect of gamma mismatch can be seen in the zone plate, where the frequency approaches 1/f the average surface goes darker... It should ideally be mid gray all the way to the edge. If you back off a few meters from the screen while the image is unscaled, you see what it ideally should look like.
Typical other areas of detail loss / underattenuation are bright areas. Look at the shirts and ties, and at the bright pink/orange flowers - and the bright roof on the red buildings.
PS: No local adjustment in the scaling, just globally applied corrections. Scaling wasn't any fancy interpolation scheme, just a integer-scale area average (with blur r0.3 before and sharpening after scaling)
DS
carstenw wrote:
Toothwalker, I see that there is a problem with the last one, but a couple of the first five look soft in your example, and sharp in James' example (maybe a little too sharp, in fact). I think it isn't necessary to be so afraid of a little false detail that you leave things soft.
That is just my preference. I am not too fond of bitingly sharp images for the web. I might have sharpened it a bit more, but decided to apply the same procedure that I have been using for the past months. The preservation of original detail is determined by the downscale algorithm.
Anyway, the posted versions that look sharper, also have noticeable aliasing in the white roof with line patterns running in wrong directions. Nothing too disturbing, but noticeable.
Toothwalker wrote:
Of course, after all this pain I expect you to demonstrate that you need 16 bits.
This is the easy part.
All of this comes to basics of processing images "naturally" or "unnaturally". In "nature" light and these things are linear. In computer due to using CRT tubes in past we have ended up to using gamma to display images correctly. However all formulas in world are linear unless they have been "compensated"/"corrected" to take the gamma into account. However in graphics almost all software does the tricks without this "compensation"/"correction" while most of the people only process only gamma raped images. Weird, but many things man have created are weird and not very logical... BTW. Also if you work with tools like dcraw, you notice that our camera sensors also output the data in linear (=gamma 1.0) mode.
I asked color managed and 16-bit in order to process images in controlled manner with "civilized" and commonly used tools (e.g. PhotoShop). However 16-bit source image is not color managed so I have to use very basic and very hard to use tools (for common population). Commonly used tools (e.g. PhotoShop) don't have good and accurate ways to adjust gamma unless you do it via color management (well considering the level of average software user that is just a good thing...). Having image in 16-bit allows one to do translations between ICC-profiles (or different gammas) without losing image quality, this is usually causing color and other artifacts and data loss when done in 8-bit.
Due to not having color managed images, all processing is done in ImageMagick command line tools - it's free tool and can be used in all common platforms (and converted from source code to less common platforms): http://www.imagemagick.org
If we look at ring chart and the differences to it when lanczos scaling is done to it:
1. Gamma 1.0 (or thereabouts, I assumed you had your non-color managed image in gamma 2.2)
2. Gamma 2.2 (or thereabouts)
ImageMagick script to produce these (after "#" character are the comments don't copy paste them if you do the same):
1.
# first removing gamma, assumed gamma was 2.2, so used value 0.45454545, which is 1/2.2
convert multipanel16bit-justring.tif -gamma 0.45454545 multipanel16bit-justring-gamma1.0.tif
# then doing resizing to 1/3 using Lanczos filter in resize operation
convert multipanel16bit-justring-gamma1.0.tif -define filter:filter=Lanczos -resize '33.333333%' multipanel16bit-justring-gamma1.0_resized.tif
# then converting image back to gamma 2.2 so that non-color managed systems show it roughly right
convert multipanel16bit-justring-gamma1.0_resized.tif -gamma 2.2 multipanel16bit-justring-gamma1.0backTo2.2_resized.tif
# saving to JPG
convert multipanel16bit-justring-gamma1.0backTo2.2_resized.tif -strip -quality 99 multipanel16bit-justring-gamma1.0processed.jpg
And if then take a look at Your lanczos (the 2nd one, without sharpening) the same error seen in ring chart as example above, creating this "box" in center and darkening the surroundings:
Here is the same with ImageMagick and using gamma 1.0 (or thereabouts) to process the whole image (Lanczos resizing like in your image - Matlab Lanczos may be better than ImageMagick, I don't have any idea of Lanczos version in ImageMagick):
So if one's preference is not to have artifacts I would recommend doing processing in gamma 1.0. Whatever script you used to create before will not work that well, for example unsharp mask filter (in PhotoShop or ImageMagick or whatever tool) acts differently. Also it's possible to utilize the gamma errors caused by processing in gamma 2.5 for example to create certain look to images.
I find the selection of 4 of 5 photos in panels is done using only one goal in mind: finding artifacts of sharpened images. If I would shoot this kind of stuff it would be damn same what sharpening is used as long as it won't cause artifacts. The only one photo had some depth of field, but it's blurry (original) and there isn't much to talk about the transition from focus to bokeh. Past years I have been evaluating sharpening and resize scripts based on "do they look the same as large non-resized or much less resized images" and "will I get same feeling of looking web photo as looking the real thing". The most important area I have been working with has been how depth of field is rendered in websize image. Typically "resize and then sharpen" will result to that the apparent DOF is multiple stops larger than when viewed in proper size. When I still shoot with "resize and then sharpen" I most often posted photos shoot f/1.4-2 to web, but when I printed same image I used version shoot at f/2.8-4. However at that time I didn't use (most probably haven't even heard of) Lanczos, which seems very good method if one prefers to do "resize then sharpen"-method. Now I have post processed only with step sharpening methods and I post same apertures to web as I print, sure the DOF get larger with step sharpening but apparent DOF is most of the times very close to real image.
I tried few of my scripts to your image, but none of them work since made for images, which are 5600px on long side and resizing them to 975px (or 972px recently due to Fred Miranda forum bug). This is 1:5.9 resizing ration, while in your example image we are using 1:3 ratio. Also the original is sharpened, and having obvious sharpening caused artifacts, which my original images won't have.
