BIT ALLOCATION IN PIECEWISE-PLANAR REPRESENTATION OF IMAGES

It is customary to represent an analog image in digital form by dividi ng its support into pixels, and within each pixel to represent the bri ghtness by a quantized scalar, i.e., to approximate the 2-D image func tion by a horizontal plane. A representation scheme in which the image function is approximated within a pixel by a sloped plane, given in t erms of the average and the slopes in both spatial coordinates, is con sidered. The problem of optimal bit allocation for the mean and the sl ope coefficients is studied for deterministic images and for samples o f 2-D random processes with known autocorrelation functions. Similar r esults are available for the 1-D case. (C) 1995 Academic Press, Inc.