计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (21): 63-66.

• 学术探讨 • 上一篇    下一篇

改进的基于MRF原理的块效应消除算法

徐培来1,2,杨 进1,刘建波1

  

  1. 1.中国科学院 中国遥感卫星地面站重点实验室,北京 100086
    2.中国科学院 研究生院,北京 100039
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-07-21 发布日期:2007-07-21
  • 通讯作者: 徐培来

Improved de-blocking algorithm based on MRF theory

XU Pei-lai1,2,YANG Jin1,LIU Jian-bo1   

  1. 1.China Remote Sensing Satellite Ground Station,CAS,Beijing 100086,China
    2.Graduate School of Chinese Academy of Sciences,Beijing 100039,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-07-21 Published:2007-07-21
  • Contact: XU Pei-lai

摘要: 基于块的DCT变换是现在应用最广泛的图像与视频压缩变换方法之一,在很多具有重要影响力的图像压缩标准如MPEG、JPEG、H.261中得到应用。通过BDCT变换,可以获得较高的图像压缩率,但是BDCT变换也会损害图像质量,其中块效应就是常见的现象。现存的块效应消除算法或模糊了图像的细节,或达不到较理想的块效应消除效果。在兼顾消除块效应和保护图像细节部分的原则上,将分区思想应用于基于图像马尔可夫随机场(MRF)原理的块效应消除算法中,提出了针对图像边界、纹理区域、平坦区域以及分块边界选择不同的MRF势函数的方法。同时,根据图像的块效应产生的灰度值偏差分布的统计分析结果改进了MRF模型的求解算法,简化了求解过程。对已有的边界检测和图像分区算法进行了改进,应用快速算法,简化了其中对块效应消除结果影响很小的计算过程。实验结果表明,这种算法既保护了图像的细节信息,又达到了良好的消除块效应的效果,并且具有较好的时间效率。

关键词: 块效应, 马尔可夫随机场, 离散余弦变换, 梯度均方差, 峰值信噪比, 图像空间分辨力, JPEG标准, ,

Abstract: Block DCT is the mostly popular transform method for image and video compression now,applied in many image compression standards with important influence such as MPEG,JPEG,H.261,and so on.With BDCT method,higher compression ratios of image are achieved.But BDCT also brings some effects that will reduce the quality of images,and blocking artifacts are usually phenomena.Many existent de-blocking algorithm either blurred images or not achieved ideal result of de-blocking.Considered all of de-blocking and protection the information of high frequency,this paper presents a de-blocking algorithm with segmentation idea into MRF,by selecting different potential function in MRF according to different area:image edge,texture area,monotone area and block edge.We also have worked on the statistics of gray offset introduced by blocking artifacts and then have improved the computing method of MRF model,which greatly simples the compute process.Haven improved the methods of edge detection and image segmentation with quickly process.Experiments demonstrate that this algorithm can achieve a good result of the protection of high frequency information as well as better reduction of blocking artifacts,and also with little process time.

Key words: blocking artifacts, MRF, DCT, MSDS, PSNR, ISR, JPEG