基于图像纹理特征的JPEG-XR帧内预测技术

Enhancing the intra-prediction in JPEG-XR by using texture information

为了进一步提高压缩效率以及重构图像的质量,使之适用于无线多媒体传感网络的要求.在讨论JPEG-XR编码预测的过程后,提出一种以图像纹理为依据的预测改进方法.该方法通过统计图像灰度共生矩阵像素间的相关特性,准确表征图像纹理的趋向,并依据纹理信息实现JPEG-XR 3个子带的预测.实验结果表明,该方法与标准JPEG-XR方法在同码率下,峰值信噪比最大可提高1.1dB,差值图像熵平均降低0.441 7.

As an emerging image coding standard,JPEG-XR can compress image in hight image quality with low-complexity,so it will be the most proposing image compress standard.To further improve the compression efficiency to meet the requirements of the wireless multimedia sensor networks.After introduce the prediction in JPEG-XR standard,this paper propose a prediction technique based on image texture feature.The method can accurate characterization of an image texture by counting relativity between pixels,and get a new method to predict JPEG-XR＇s three subbands according to the statistics.The experiment results object evaluate of our modified JPEG-XR scheme works better than the JPEG-XR standard both in bit rate and reconstruction image quality.Compare with the standard JPEG-XR,our method got 1.1dB increase in PSNR under the same bit rate and the average entropy of difference image decreased 0.4417.So it is an promising method in JPEG-XR image processing.