MSE and PSNR prediction for ADCT coder applied to lossy image compression

Abstract

Lossy compression of images is used for numerous applications nowadays. Typical requirements to it are to provide a higher compression ratio (CR) for a desired quality and to perform this quickly enough. Advanced discrete cosine transform (ADCT) based coder potentially provides a good compromise between CR and image quality but consumes essential resources for reaching a desired quality of compressed data. To get around this shortcoming, we develop a fast and rather accurate approach to prediction of mean square error (MSE) or peak signal-to-noise ratio (PSNR). This approach performs sufficiently faster than compression and allows saving time and resources. It is shown that if an image to be compressed is corrupted by noise, prediction correction is possible and desirable.