The sentinel

[peripatetic]

Thanks for all the comments.

I don't think there was a dynamic range problem in this instance, the histogram is pretty clean, no clipping at the top and most of the bottom stop wasn't used at all.

It was pretty humid, the mountains having been covered with cloud, mist and rain the day before and after. So the haze is due to the conditions.

A polarising filter would have been a good idea, unfortunately the lens I'm using suffers from dreadful vignetting at the wide end of the zoom if any kind of filter is used, so I don't use filters on it. This shot was at the telephoto end so it would have been useful for sure. There are often real-world restrictions on these things though; I was walking with my wife, and although she doesn't mind me taking photos, she doesn't stop and wait either. Swapping lenses and putting filters on and off isn't something easily or safely done while walking on rugged trails. Win some, lose some. :-)

As to precise colour and saturation - this was slightly tricky for me because I'm not used to processing JPG, but I'd switched over to JPG because I was running low on space on my CF.

I know also that the print will have bolder looking colours than the web output. I agree that the extra contrast helps the small shot, but the contrast boost seems to lose detail at higher resolutions, so I'll just have to try to optimise for the print.

(I'll be back in the UK next week and will have access to my printer again!)

The colours in the version I posted are pretty close to what I remember seeing, the sky was a fairly pale blue, the clouds were just beginning to break up, and I love the way they were rolling around the summit.

[Aumma45]

Peri, you are a very good critique of yourself. That is the spirit. Regards. Jaki

[squirl033]

peripatetic wrote:

There are often real-world restrictions on these things though; I was walking with my wife, and although she doesn't mind me taking photos, she doesn't stop and wait either. Swapping lenses and putting filters on and off isn't something easily or safely done while walking on rugged trails. Win some, lose some. :-)

i know that feeling! :lol:fortunately, i don't need to swap lenses, so there's at least one small advantage to my FZ20... i do like this shot, though.wish i lived someplace where i could get pics like this...

[tclune]

Hards80 wrote:

its not dynamic range that is the problem here, its the haze from atmospheric conditions.. and not a whole heck of a lot you can do about that..

i think one of the reasons you might be seeing it as cutout looking is that i think this was taking at a moderate to telephoto focal length, which tends to stack up objects and lessen the apparent difference..

You may be right. But I have worked professionally for many years with image processing (admittedly in the very highly specialized context of medical imaging, but much of what I've learned has general applicability). All digital cameras process images. One thing that can happen when you have highly bimodal images with limited expression of thefrequencieswithin the two histogram peaks is that the sharpening over-accentuates edges in a way that makes them seem artificial. You will often see a Gibbs effect ("ringing" at sharp edges) if you graph a single raster line that cuts across the offending area. I have not actually performed that experiment on this image, but it certainlly has the look of this effect.

You can get the camera's internal algorithm to behave itself in these circumstances by various methods. First, you canlower the "sharpness" or "contrast" settings on the camera if it offers you that entry point. Most cameras do. Second, you can use a GND, although the extent to which that would help in this particular case is limited because the Sentinel juts into the region that you would naturally be toning down, so the contrast with the sky on either side of it would not be helped. I think it would be more effective in dealing with the same, although more muted, effect between the foreground bluff and the valley leading up to the Sentinel.

My sense is that photographers tend to overlook the effects of these kinds of issues, which is why I raised the point in the first place. Of course, I may still be off base in this case.

[peripatetic]

I'm intrigued now...

I usually shoot RAW, but not long before this shot I switched to JPG because I was starting to run low on space on my CF card.

I was somewhat disappointed in the image quality here.

Is the histogram useful here?

What else could you use to decide?

[peripatetic]

Here is a 100% crop from the image.

I'm pretty sure I have my JPG settings set for very conservative sharpening - you can always add more but it's hard to take away.

[tclune]

peripatetic wrote:

Is the histogram useful here?

What else could you use to decide?

Yes, the histogram has value. But, for what I am thinking of, the most useful diagnostic is to graph a line or column of luminosity data. Just plot each point on a graph as it appears in sequence. What you get is a graph of the grayscale transition between adjacent pixels. (You can do this for the color component if you like, but usually the luminosity graph will tell you anything that you need to know.)

What you'll see in the case of extreme sharpening is a ripple in the line around a sharp edge. It may just dip down before it shoots up, or may even dip up and down at one side of the sharp edge before it shoots up at the other, and ripples again at the top of that. In either case, the mathematical term for this dip is the "Gibbs effect." If you have an infinite number of frequencies to reconstruct the sharp transition, you get a line that overshoots the top of the square wave. If you have a truncated series of freequencies (as in the case of a digital camera), you get a more gentle dip down before the sharp rise (which never quite manages to go vertical), and then there is a hump at the top of the rise before it settles back down to the new height. This kind of thing is a graph of oversharpening. The more that you have higher frequencies present, the less pronounced the "mistracking" of the sharp transition -- detail in the image around the sharp edges limits the apparent effect (but the basic underlying circumstance is an inevitable mathematical limitation of trying to track a singularity. You can just limit the visual impact of it.)

Unfortunately, just as most folks prefer almost cartoonish color saturation, most folks also prefer cartoonishly over-sharpened edges. If you look at the professional photographs offered at most vacation spots, they are all sharpened to a point that I find physically uncomfortable to view. I confess I don't quite "get" the normal taste for such things. This image does not have anything like that level of problem. But it appears to me to be there to some extent. My tastes are certainly not universally shared, so please take my comments for what they are worth. And I do want to reiterate that I like the composition very much. I just would prefer more "natural" transitions around the sharp edges. Of course, as always, YMMV.

ETA: I have sometimes experienced this sensation when viewing nature, which I presume is due to an artifact of my visual system (I have rather severe astigmatism). So I would suggest that you may want to pay attention to the original scene, and if the photo captures faithfully what you have seen, it's not a problem in the camera's algorithm. These things are always tricky...

[peripatetic]

The more that you have higher frequencies present, the less pronounced the "mistracking" of the sharp transition -- detail in the image around the sharp edges limits the apparent effect (but the basic underlying circumstance is an inevitable mathematical limitation of trying to track a singularity. You can just limit the visual impact of it.)

All new to me, but fascinating...

Would you then expect to see more of these artifacts from 8-bit (JPG) images than 12-bit (RAW) images converted to 16-bit TIF files? (Please forgive me if I have grasped the wrong end of the stick here - I'm quite happy to let go. :-) )

In this particular image though it was converted to JPG in the camera, but my parameters for sharpening, saturation and contrast were all at the relatively conservative defaults for the 20D.

As a general principle the last thing I do, after all my editing (if I ever need to use JPG) is to convert to 8-bit mode.

Unfortunately, just as most folks prefer almost cartoonish color saturation, most folks also prefer cartoonishly over-sharpened edges.

And too much contrast for my liking too. One of the things I really like about the Canon DSLRs is the anti-moire filter reduces acutance in the image, without apparently reducing resolution much. It's essentially a soft-focus filter. I tend to take my sharpening, contrast and saturation on a case-by-case basis, sometimes it's just the wrong thing to do for an image.

I must admit to finding the desparate attachment to "sharpness" as the overriding quality a lens must possess to be rather tiresome. There are so many other factors to consider.

So I would suggest that you may want to pay attention to the original scene, and if the photo captures faithfully what you have seen, it's not a problem in the camera's algorithm.

Indeed, this is where the photographer has the advantage over the viewer. As it happens these are very sharp mountains. Nature seems to have over-sharpened them somewhat. :lol: The image I posted originally has the closest fidelity to my memory of the actual scene. (Though thanks to everyone for their edits; it is through experimentation that we learn.)

I'll be posting a few more images of the day here in the next week or so, as I get a chance to work on them, so you will be able to see what I mean.

[tclune]

peripatetic wrote:

Would you then expect to see more of these artifacts from 8-bit (JPG) images than 12-bit (RAW) images converted to 16-bit TIF files?

Yes. The less processing, the less Gibbs effect. Furthermore, the JPEG suffers from a double-whammy (to use the technical term...) By dropping to 8 bits, fine gradations are lost and the quantization fo the JPEG on the 8-bit data further accentuates the loss. If the only high-frequency data left is the sharp edge, it will have an accentuated edge effect, which is just plain harder to track than if you also have fine detail around the sharp edge -- the net effect of the added frequencies is to "fudge" the edge, allowing for higher fidelity in the algorithm. I once was bitten by this very circumstance in an algorithm I developed for enhancing scintigraphy images (nuclear medicine images using radioactive decay). I created an algorithm using something called Weiner filters that did a wonderful job on a wide variety of images. When we were confident enough to go into testing with it, one of the first patients had had their thyroid removed surgically. The radioactivity profile that resulted was of a tiny sharp region of very high counts in a sea of very low counts. The Gibbs effect was very pronounced. In this case, the shadow that my algorithm created would have been interpreted as a return of cancer, and very possibly led to an unnecessary operation. In a thyroid -- even a diseased thyroid --where there was more fine structure, we never hadexperienced this error in the processing routine. Deliver us from the Gibbs effect!

[Hards80]

Tclune.

thanks for the technical information. i am familiar with the concepts, more specifically, the results of these concepts. so it is really nice to have a little more background information regarding these phenomenon. very interesting read.

-dustin