Blockoverlap transform coding of image signals
Abstract
Blockoverlap decorrelation methods for transform coding of image signals are discussed theoretically. Methods which require the least computing time are considered, including transformation with overlapping time windows and fast convolution transformation. In an examination of system algorithms it is found that with overlapping time signal transformation, signal reconstruction is not completely exact because the intermediate spectral coefficients are suppressed during transformation. This problem can be alleviated by making the system of functions completely orthogonal through bilateral quadratic transformation. Fast convolution has the drawback of having to transmit two coefficients for each coded signal value. The additional coefficients are redundant and can be suppressed, because at the decoding end, an equal number of signal values in each segment is already known prior to transmission. A series of schematic representations of the systems is provided.
 Publication:

Siemens Forschungs und Entwicklungsberichte
 Pub Date:
 1984
 Bibcode:
 1984SiFoE..13..100S
 Keywords:

 Image Processing;
 Signal Encoding;
 Blocks;
 Convolution Integrals;
 Fourier Transformation;
 Redundancy Encoding;
 Communications and Radar