Augmenting JPEG2000 With Wavelet Coefficient Prediction

This paper presents a novel modification to the Discrete Wavelet Transform used in the JPEG2000 standard. The modification utilises a learning-based method to predict the detail coefficients from the approximation coefficients at each level of the transform, allowing for higher compression rates while maintaining most of JPEG2000's appealing features. A coefficient prediction method based on single-image super resolution literature has been implemented and tested to assess the potential of the compression system. We have observed that the accuracy of the prediction is highly dependent on the content of the test image, with certain features being highly predictable and others not predictable at all. In the best performing test images, we see a quality gain of up to 1.5dB or equivalently a reduction in bit-rate of up to 14% over the JPEG2000 standard.

1. Introduction

The JPEG2000 compression standard not only offers high coding efficiency but also supports several additional features that other state of the art formats do not. These features include scalability, optimised rate control without iterative coding and region-of-interest accessibility [1]. These features make it particularly desirable for applications that involve access of images that are extremely large by allowing not only efficient and well-controlled compression but also quick access to specific regions of the whole image. From a pure performance stand-point however, JPEG2000 is typically inferior to current state of the art methods, especially those based on H.265/HEVC such as BPG and HEIF.