An efficient lapped orthogonal transform image coding technique

Abstract

A fast and computationally less complex coding techque is described which uses partial-LOT computation a l g o r i b and efficiently cfiscards perceptually insignificant high frequency transform coefficients. The coding process involves AC energy classification, human visual system weighted normalization and quantization. The values of normalization factors are image independent and governed only by the bit-rate and activity index of the image blocks. Zones of various shapes and sizes are chosen for each activity class to perfom zonal sampling which efficiently cfiscards high frequency coefficients having zero values. A progressive transmission version of the proposed technique is also given. This technique gives better performance at comparable bit-rate than that obtained by using LOTlCVQ and LOTNQ schemes.