AUDIO PITCH DIRECTSHOW FILTER

A200/Tamron 18-250 Hoya MCUV filter, ISO200 > dcraw > Gimp with gradient blends & usm]



This was my original "blend". I started by brightening the center +80 and the contrast +75, the part away from center by about +40/+30 & blending with a bidirectional gradient in the middle at about 70% opacity. It came out slightly dark on the right in the middle yet with insufficient contrast away from center...

...it would be very-nice to be able to adjust the two layers and the mask independently and then see a combined version of them in real-time...

And to run a split mask in the middle and reduce the contrast on the right, but after a few attempts at that I gave up as it didn't appear to be worth the effort. Like I said it would be better to see these things before blending rather than after, I can only spend so much time & effort blending two layers on one image for Flickr. Just imagine doing this for an entire video. Literally I was just thinking of writing a 2D low-pass filter for this. Anyway on the bright layer you want to get the contrast in the ballpark, a little low and get the brightness a little high because the blend is going to darken and increase the contrast of the light parts and of course adding more contrast after the blend will make them that much darker and blow-out the highlights & shadows that much more. But the point is that with insufficient contrast the midtones will look pasty and gray and the whole shot looks fake. Obviously that can be overdone to the point where it is overcontrasty and fake, as well. So it's a balance.

Try starting with a linear shot, a raw conversion with no contrast added and the gamma set to 1 ;)

Or try it with two different exposures of the same shot, say one generated with H0W and H7W (with dcraw) and then use a threshold mask and an amorphous gradient blend ;)

This is begging to be done with a "smart" filter, really. But it's why I think the best approach is to just shoot it to expose the highlights properly and then rescue the shadows with a gradient blend, not use a complicated HDR tool and multiple exposures. If it's shot so that the highlights are slightly blown-out but the shadows are better-exposed, and then the shadows are rescued with H7W in dcraw (or the equivalent in your favorite RC) and you try to blend an H0/7W exposure set, then the shot will be slower and less stable. Plus that's a fine line to walk while you're in the field. It's easy to look at the LCD and see if the sky is blown-out, if not all the highlights. That's where I would start, maybe minus half a stop or so, as there will be saturated sections that are not "highlights", like in the above where the sky is saturated red in places. Pushing those will change the color entirely. Oversaturation is another reason why I don't like to shoot to the right even if the camera is noisy in the shadows, and you shouldn't need a meter to tell you what is happening to the highly-saturated parts of the scene if the scene has over 3 stops of DR and the overall exposure is near 0eV according to the camera meter. It is all a balance between the advantages of a certain technique and its negative consequences, in this case you have to balance blowing-out the highlights and getting a flat, dull result with getting noisy shadows and midtones and a shot that is full of crud. So this is one more reason to bracket even if I'm going to end-up throwing-away a third of the shots right there and another 3rd or more once I get home. And obviously if more than one exposure is used and it's not perfectly-aligned, then the edges will get softened by the blending process. So I always work from just one exposure and now I clone it to different layers, push them around, add contrast and blend. You can go from there to doing selective editing on each layer.

The problem of course is the idea of doing all this editing on thousands of shots, even if it can all be done fairly-easily by hand, which in general is far from the case. It simply is not easy to automate vignetting, CA & geometry-correction to the point where it is well-optimized for each shot. Not to mention if you don't hold the camera properly or crop or rotate the shots before processing, not to mention correcting the lighting and doing NR if you want to go that far. It *is* however easy to rip-off a half-a dozen thousand shots with average post-processing especially if you're happy just shooting jpegs. And in the long run, guess which really matters the most: excellent IQ on a handful of shots or good average performance over years of use. Especially when, let's face it: "excellent", for digital photography, means "a decent representation of what we would see if we were there". You're just not going to match, much less "beat", a view of the real thing with a camera, much less a digital camera. If you're lucky it'll be "decent". Sometimes even "impressive&qu

MeeBlip running with an RC low pass filter , two 10uF caps.