I am on my way now to start making a simple image processing program which shall have function such as blurring, and ... well, other image manipulating functions. But I am not sure whether I shal use an array with unsigned ints to store the whole RGB value, or one array with signed ints for R, one for G and one for B. I believe that using one array for the whole RGB value will be faster, especially since if I use SDL I can write directly into the buffer and display it directly, but using one array for each layer will be much more precise. Do you think there will be visible losses in quality after a few manipulations if I choose to use the less precise type? Which of these methods do you think I should use? What does other raster graphics editors use, like Photoshop and GIMP?
I also intended to build a simple computer vision library with a few functions, maybe these are not as picky as a human.