Re: [Digital BW] Grain and Photoshop
2005-05-24 by hogarth@snappydsl.net
I'm not sure there is a good answer for this. First, some observation. To get 4-6 pixels per film grain clump, either your grain is just huge, or you found a drum scanner with a one micron aperture. IIRC (and it's been a long time -- I'm sure someone will correct this if it's wrong or out of date) film grain clumps usually range from about three microns on the small side up to about 12 microns on the huge side with the median being closer to about six microns. And if you are getting 4-6 pixels per grain clump, you are getting a huge amount of pixels that contain the "nothing" that is the film base *between* the grain clumps. IOW, it must be a really big enlargement. IOW, the Photoshop people probably didn't plan on this happening a lot. This brings us to the idea of how to deal with it. Since film graininess is at least partly a function of density, one would want the software to grow and/or shrink the grain size as you change the "density" of the image via the levels and curves (etc.) tools. This would be difficult - complex mathematics, probably an order of magnitude more complex than unsharp masking, for example. This kind of manipulation would also require Photoshop to have some knowledge of the film itself. Currently, Photoshop has no information on the image source. To get this information, someone somewhere is going to have to test fairly extensively the films of most interest. At the very least, a color transparency, a color negative, and a B&W negative. Then people will want their own films. Then people will want their own films with their own developers. I think it would become irreducibly complex. IOW, I think it safe to say that Photoshop will never handle this. So.... how do we handle it with the tools that we have at the moment? I think the best we can do is to try to preserve the size, shapes, and locations of the grains as best we can. This means some capture sharpening to help overcome the softness of scanning, and then some output sharpness to help overcome the softness of the output device (for example, an inkjet printer printing at 2880 uses 8 ink dots [with the attendent dot gain] to make a single pixel at 360ppi output). We also need a good scan to avoid the need for heavy manipulations. This will minimize the problems of mismatch between grain size and image darkness/lightness. Not much help is it? Sigh... I think it all comes back to the "basics" of needing an exceedingly high quality scan, manipulating the image as little as possible, and careful use of sharpening. These are the same things we generally do for every image though. I wish I could offer more help. -- Bruce Watson Ernst Dinkla wrote:
Show quoted textHide quoted text
> Better minds than mine must have thought about the following long > before I started wondering: > > A lot of the functions in Photoshop rely on the principle that color > and greyscale are represented in the pixel information and the > distribution of the pixels. When grain is resolved in color or B&W > above the level of 1 pixel per grain say at 4 or 6 pixels per grain > there's another method of color and tone representation interfering > with the pixel information. More extreme in contrasty images with > coarse grain. Although the grain doesn't have all the properties of > the true B&W screen in offset printing and the grainsize will pair > size with density to some degree and doesn't have a hard edge, > nevertheless there's information in that irregular screen that behaves > differently to the information in the pixels it is build on. With the > result that the more contrast and the bigger the grain the less > control you have on tonality etc in Photoshop. Like having a true > bitmap on top of a greyscale. The simplest example of that observation > is when you brighten the shadows the grain there isn't getting much > smaller but becomes more grey. In the highlights the grain will change > its size as the density - size is more related. The histogram of a > grainless image and one of a grainy image can show a high contrast for > both while the grainy image isn't contrasty in general but only on its > grain level. This doesn't imply that there's no editing possible but > ignoring what happens seems impossible. Sharpening can decrease the > brightness for example. > ' > What is the way to cope with it without removing the grain ? Sure I > can judge with my eyes on a calibrated monitor but even then there are > good reasons to expect a different outcome from the print as the grain > will go through that process in another way than its rasterisation to > the screen. > > Ernst >