|
|
|
“TRANSFORM IMAGE CODING”
by William K. Pratt and Harry C. Andrews
March 1970
The basic goal of digital image coding is the development of a coding technique that permits the representation, and subsequent recovery, of an image by a minimal number of code bits [1-3]. In some applications virtually no image distortion is permitted in the coding process, while in other applications a controlled amount of distortion is allowable in the achievement of a substantial bit reduction. In general, when redundancy is removed from a data source, the compressed data is more sensitive to the effect of channel errors. One of the restrictions in selecting a coding method, therefore, is that the compressed data must not be overly sensitive to channel errors.
In 1967 a new technique of image coding, called Fourier transform coding, was developed at the University of Southern California [4-6]. Another related method, called Hadamard transform coding, was discovered at USC in 1968 [7-8]. Since then investigations have been made into the applications of other mathematical transforms for image coding. Out of these studies has emerged the generalized technique of transform image coding [9-11].