Methods for quantizing and compressing digital image data

摘要

There are disclosed methods for quantizing image data with signal-dependent noise to a fewer number of values so that noise information is discarded while signal information is maintained. This is achieved by deriving a set of equations, the solution to which may be numerically computed and whose solution constitutes an optimal relative mean square error quantization for a signal with noise. Two low complexity explicit heuristic formulas for quantization are also provided. Also disclosed are methods for compressing image files containing signal dependent noise by means of near lossless compression so as to maintain signal information in the compressed data while discarding noise information, and methods for compressing multi-component color image data so that psycho-visually significant color information is maintained in the compressed data while other color information is discarded and a method for compressing multi-component color image data which is linearly proportional to high dynamic range light intensities so that color information significant for display and printing is maintained in the compressed data while other color information is discarded.