Automatic lossy compression of noisy images by spiht or jpeg2000 in optimal operation point neighborhood

Abstract

It is often needed to compress images with providing certain properties or quality. Wavelet based coders SPIHT and JPEG2000 easily produce a desired compression ratio but not quality especially if one deals with compressing images corrupted by noise for which specific behavior of quality metrics on compression ratio (CR) or bpp might be observed. In particular, optimal operation point (OOP) where a metric determined for compressed and noise free images might have optimum can take place. Lossy compression in OOP or its neighborhood has several advantages, However, for its attaining for such wavelet based coders as SPIHT or JPEG2000, only iterative procedures that can be rather time consuming have been proposed so far. Here, we propose a single-step procedure for determining bpp to be set for providing compression in OOP neighborhood. This procedure exploits two features. First, rate/distortion curves for SPIHT and JPEG2000 are shown to be very similar to those ones for DCT-based coder AGU. Second, a fast and simple procedure for predicting compression ratio in OOP neighborhood (without image compressing) has been proposed for the coder AGU recently. Examples demonstrating that the proposed procedure performs accurately enough are given.