Joint coding method for images and videos with multiple arbitrarily shaped segments or objects

Abstract

An adaptive image coding method and system are disclosed. The system accepts an input image, divides it into image segments, and assigns each segment to a wavelet transform filter from a bank of such filters for transformation. The bank preferably comprises filters adapted for different types of image content, e.g., sharp edges, slowly-varying contours, etc. Filters are preferably assigned so as to produce minimal distortion for their assigned image segments at a given target bit rate, each filter produces transform coefficients for its segment using scale and subband settings common to the entire image. The valid coefficients for each segment are then combined in a composite wavelet coefficient image, which resembles a single wavelet transform of an entire image—although different filters are used to create different portions of the coefficient image. The composite image allows joint, rate-distortion optimized coding for a segmented image. Joint coding allocates bits between the transforms of the image segments optimally, and produces an easily scaleable bitstream.