Distortion-Controlled Dithering with Reduced Recompression Rate

Morriel Kasher,Michael Tinston,Predrag Spasojevic

2024-02-26

Abstract:Dithering is a technique that can improve human perception of low-resolution data by reducing quantization artifacts. In this work we formalize and analytically justify two metrics for quantization artifact prominence, using them to design a novel dithering method for distortion-controlled data compression. We present theoretical entropy calculations for this dither and experimentally validate its performance on a low-rate image compression task. The result is a drastic improvement in the perceptual quality of quantized images with a lower recompression entropy than any state-of-the-art dither technique, achieving 45 points lower PIQUE at the same rate or 40% lower rate at the same PIQUE. The proposed dither is an adaptable tool applicable for use in any lossy compression system, permitting precise control of rate-distortion characteristics for both compression and recompression.

Signal Processing

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in low - resolution data compression, how to reduce quantization artifacts through dithering techniques, thereby improving human - perceived quality and reducing the recompression rate. Specifically, the author proposes a new dithering method, aiming to precisely control distortion characteristics in order to achieve better perceived quality and lower entropy.

### Main problems and goals

1. **Reduce quantization artifacts**: Traditional quantization methods will produce obvious artifacts at low resolutions, affecting the quality of images or audio. The paper proposes to reduce these artifacts through dithering techniques, thereby improving human - perceived quality.

2. **Reduce the recompression rate**: In many applications, data needs to be decompressed first and then recompressed for transmission or storage. One of the goals of the paper is to design a dithering method that can reduce the bit rate required for recompression while maintaining high quality.

3. **Theoretical basis and optimization**: Existing dithering techniques lack strict mathematical theoretical support and do not provide practical design tools. The paper attempts to fill this gap by introducing new distortion metrics, optimizing the dithering distribution, and verifying its performance.

### Specific contributions

- **New distortion metrics**: The author proposes two new distortion metrics, namely metrics based on absolute error autocorrelation (AR1) and mean - square - error autocorrelation (AR2), for evaluating the perceived impact of quantization artifacts.

- **Parameterized dithering distribution**: Two parameterized dithering distributions are designed, optimized for different distortion metrics respectively. These distributions can be flexibly adjusted in different application scenarios to achieve the best compression effect.

- **Experimental verification**: Through theoretical analysis and experimental verification, the superiority of the proposed dithering method in low - bit - rate image compression tasks is proved, especially outstanding in recompression.

### Summary

By introducing new distortion metrics and parameterized dithering distributions, this paper solves the problems existing in existing dithering techniques in low - resolution data compression, significantly improving the perceived quality and reducing the recompression rate. This provides important theoretical and technical support for future image, audio and other types of compression.