Digital filters implementation using DPCM video signal processing

Abstract

A new realisation of DPCM video signal/image processing is introduced to increase the throughput rate and reduce the hardware cost of the 2D systolic digital filters. This achieved by using a DPCM coding system with a 2D predictor and a 27-level symmetric nonuniform quantiser prior to processing. The advantage of the symmetric nonuniform quantiser is that the size of memory required for multiplication is reduced by a factor of 16 compared to 4 in the existing DPCM implementations. It is shown that the new realisation results in a 64% increase in the throughput rate and a significant reduction in the hardware cost. From the objective and subjective tests carried out, it is shown that the new realisation results in images with visual quality similar to those obtained using conventional processing. Also, the effect of the choice of the 2D predictor is discussed, where it is shown that a 2D predictor with integer coefficients is the most appropriate for DPCM processing.