Well, just when I have defined all contents for my book (I can't seem to see the end), I find several posts related to the "Pseudogrey" concept.
Normal grayscale images are coded in 8 bits in a single channel (one byte per pixel). This way, up to 256 different levels (or shades) can be represented, from value zero (pure black) to pure white (value 255).
That amount of shades is usually enough for most applications. However, in some specific cases this is not enough. For example, in a picture with an even sky we can see banding effects or patterns, a similar effect to posterization in colour. Generated contours are visible, and the outcome may not look nice or natural.
Solution is, of course, to increase the number of bits to define each colour, making transitions softer. Full process would be as follows:
- 8-bit image is transforned in a RGB, 16-bit-per-channel one (only 12 bits to be used).
- bits are moved 4 positions, so that the original 8-bit information is "shifted-to-right"
- "1" is added to one or two channels (or none!) depending on the values of the lower nibble (4-bits group) in the original data.
Output file has up to 1786 different values (we started with just 256). Thus it will also not use the full potential of 12-bits (up to 4096 colours - that would make "deep grey")...
As an introduction, I think this is enough for the moment; I will investigate how to do this with GIMP to produce this kind of file. To begin with, JPG file will not be valid anymore due to its limitations (number of bits per channel, eight - we would trim colours back). But I already found some webs where information is provided... for example: