A fast adaptive cosine transform image coding technique

A computationally efficient adaptive cosine transform coding technique is described in which only 10 x 10 DCT coefficients are required to be computed for a 16 x 16 pixel block. The coding process involves fuzzy based ac classification along with normalisation and quantisation which are governed by the activity index of the blocks. Adaptive block distortion equalisation is also performed on the blocks to eliminate block edge structures. Zones of various shapes and sizes have been chosen for each activity class to perform zonal sampling. Very good reconstructed image quality is demonstrated for coding of gray level test images at a bit-rate of less than 0.3 bit/pixel which is lower than most of the other adaptive DCT techniques.