Wavelet-based image compression using perceptual distortion metric

In this paper, we have implemented a dithered quantization model to be used with perceptual distortion metric. Mean-square-error (MSE) based cost functions are mostly used for image coding. However, MSE does not necessarily preserve the perceptual quality of the original image. We are using Minkowsky metric as the criterion for allocating bits into various subbands to achieve desired degree of compression. However, conventional deterministic quantization fails to preserve certain precondition required by perceptual coding using Minkowsky metric. In this paper, we have used a dithered quantization model to remove the limitation imposed by deterministic quantization. Experimental results show improvement in perceptual quality over the conventional deterministic method based compression system.