Photomatix: A Virtual Magic Wand

1. As with most other functions, the image result should gain if quality improvements are already implemented in the RAW data. So where is the first image processor that features real time local contrast improvement?

I suspect that the only place we shall ever see one is at perpetual_motion_machines.com. At the level of the sensor, the basic elements of a picture--pixels--do not exist. One pixel is formed by combining the information from three (Foveon) or four (Bayer) cells. Without combining cells into pixels, no “smart” contrast manipulator could “know” what information the picture contains. Thus, to apply any “smart” algorithm, this real-time manipulator would first need to make an ad hoc conversion of cells into pixels. Since this would amount to a raw conversion, the contrast enhancement would not be intrinsic to the raw data, it would be a manipulation of converted data.

Even if some processor were fast enough to do this in real time, I would not want it done. It strikes me as unlikely that any ad hoc conversion in real time could be as good as a considered one later. Moreover, if it were, manipulating the raw data within the camera would distort the data, thereby vitiating the advantage of a raw image. Even if somebody could develop a way to enhance the contrast in real time, somebody else is likely to develop a better way later, so it I would want to keep the raw image intact.

2. A question for perception physiologists: Is there a point of "supersaturation"? Looking at Charles' images I found each of them extremely stunning but I was wondering whether a slide show with 100 images with optimized local contrast might exhaust the visual perception. I believe something similar happens when we listen to loads of MP3 music that are poorly compressed or "sound-optimized"

I don’t listen to MP3 music, so I don’t know what effect you are describing, but I often compare photographs to radios. Reproducing a landscape on paper is comparable to reproducing a symphony over the radio. The dynamic range of each must be compressed dramatically while simultaneously, to appear realistic, contrasts must be enhanced. A solo flute within an orchestra will always need to be jacked up. However, the flute should never be jacked up so loud that it sounds as large as the orchestra.

If you wonder how a series of optimized images will look projected, spend an evening at your local movie theatre. In most scenes in most films, the lighting is arranged with great sophistication to optimize local contrast.

I think your two questions are related;

You state: 1. As with most other functions, the image result should gain if quality improvements are already implemented in the RAW data.

I disagree with that. I think the raw data should remain just that-raw-i would prefer compressed TIFF. I am continuing to go back to print film (I often don't get prints) for the same reason people left chrome for print film in the 80s--much greater dynamic range. This helps you when you are shooting outdoors and also when you miss the proper midrange exposure by a stop or so. But it requires judgment to take the wide dynamic range of film and compress it to low dynamic range paper. No computer can tell why you took the picture and what you want to show.

Of course autolevels works less than 20% of the time as even a first step--but it isn't just the automatic part, it is rather this question--which is more important, the girl or the trees?

2.. . . . I was wondering whether a slide show with 100 images with optimized local contrast might exhaust the visual perception. . . .

This gets right to the point. If everything is equally visible, then what is the point of the picture? Just because the second violin part is more interesting than the first, that doesn't mean it should be louder. As my mother always pointed out to me there is only one way to make a violin sound loud--play the other passages softly.



The thing I have learned is that my printer is able to do black very very well. Make use of that. Think about the Dutch Masters.

----------------

Every now and then I will get a picture in photoshop and run it through a series of improvements. 1 is better than 2; 2 is better than 3; 3 is better than 4; and so on. But sometimes, even when 8 is better than 7, when I go back--like an Escher painting, I find that 1 is better than 8. Somewhere along the line, the photo is no longer the photo and I have to go back to the original image.

so I think the RAW file should remain really raw. ------------

The same is true with software that gets rid of grain. It makes everything look like plastic.---but I use GEM and ninja looks good, too, it's just that sometimes it helps a little and sometimes not at all and sometimes it looks sick. I really like the looks of this new stuff, but I doubt it will do more than make one out of 20 steps in fixing a picture.

-------------- Even so, I think Mr. Maurer agrees with Ansel Adams: "The negative is only a tool to create the photo."

At 08:22 10/17/06 -0700, Bernhardt, Sebastian wrote:
>I was wondering whether a slide show with 100 images with optimized local
>contrast might exhaust the visual perception.

A hundred consecutive images with insufficient variety will exhaust your
perception, no matter in what sense the variety is lacking. Thus a hundred
consecutive images with optimized local contrast will be interesting only
if you are studying contrast optimization.

>I believe something similar happens when we listen to loads of MP3 music
>that are poorly compressed or "sound-optimized".

Poorly compressed MP3 sound is the equivalent of a photograph with contrast
*decreased* and noise *added* -- exactly the opposite of the sorts of
processes that Charles Maurer discussed.

I look at some of the processed images on the hdrsoft.com web site and
think, wow, that looks like a painting rather than a photograph. Of course,
this will come as no news to painters (such as my wife), who have always
adjusted the lighting in various parts of the images they create to show
more than could be seen at a single lighting level. This somewhat surreal
quality doesn't put me off -- in fact, I'm eager to find time to try out
Photomatix, and have some specific photos in mind that I'd like to try it
on -- but thinking about this may help to delineate just where such
processing is useful. Clearly it's not a process which is intended to be,
nor ought to be, applied to every image.

Edward
--
Art works by Melynda Reid: http://paleo.org