Image compression in signal-dependent noise

Abstract

The performance of an image compression scheme is affected by the presence of noise in an image. This work mainly investigates the effects of signal-dependent noise on image compression using the JPEG image compression algorithm. Simulation results show that the achievable compression is significantly reduced in the presence of noise. The types of noise considered are, signal-independent additive noise, signal-dependent film-grain noise and speckle noise. For improvement of compression ratios noisy images are pre-processed for noise suppression before applying compression. Two approaches are used for reduction of signal-dependent noise prior to compression. In one approach estimator designed specifically for a particular signal-dependent noise model is used on the noise degraded image for noise suppression. In the second approach the signal-dependent noise is transformed into signal-independent noise using a homomorphic transformation. An estimator designed for signal-independent noise is then used on the transformed image for noise suppression followed by an inverse homomorphic transformation. The performances of these two pre-compression noise suppression schemes are compared using different performance criteria. Simulation results show that pre-compression noise suppression significantly increases the amount of compression obtained subsequently. The compression results for the noiseless, noisy and restored images are compared.