Embedded color image coding using context-modeled wavelet difference reduction

A novel embedded wavelet coding algorithm for color image and video compression is proposed. A color image is first represented in YUV color space. The discrete wavelet transform of each of the three color components is then calculated separately. The luminance component is compressed with context-modeled wavelet difference reduction (CM-WDR). The chrominance components are compressed using the contextual information obtained during the coding of the luminance component. Our experimental results show that the proposed algorithm performs better than two latest wavelet-based benchmark schemes, namely CEZW and JPEG2000, in terms of color SNR and component PSNR.

1. INTRODUCTION

Most wavelet-based image coding schemes assume that the input is a grayscale (or monochrome) image. However, color images are almost always involved for real-world applications. A simple and direct extension of these schemes to color image coding is to encode the components of a color image as three independent grayscale images. This strategy is adopted by the latest still image coding standard JPEG-2000 [1]. In the JPEG-2000 coding standard, a color image, such as RGB, is first decorrelated using YCbCr or RCT transform before the wavelet transform. After the color transform stage, the decorrelated components are treated independently as grayscale images.