二维相机阵列多视点图像信号压缩方法研究  

Approaches to multiview image coding for 2D camera array

作　　者：邱涛[1] 符冉迪[1] 蒋志迪[1] 

机构地区：[1]宁波大学电路与系统研究所,浙江宁波315211

出　　处：《浙江大学学报（理学版）》2008年第2期163-167,172,共6页Journal of Zhejiang University（Science Edition）

基　　金：国家自然科学基金资助项目(60672073);教育部科学技术研究重点项目(2060059);浙江省自然科学基金资助项目(Y105577)

摘　　要：结合二维相机阵列多视点图像信号的特点，研究了二维相机阵列多视点图像压缩空间预测的不同方法．根据GOP中I帧的个数、位置等不同情况，对所给出的3种类型的多视点编码方案，9种典型的预测参考方式进行重点分析，对比了各编码方案的率失真性能，得出相应的结论．所给出的二维相机阵列多视点图像信号空间预测采用多级B帧的方案，相对于其它方案在相同比特率下峰值信噪比（PSNR）平均提高约0．2～0．8dB．Multiview image is a new-type of digital medium that can improve 3D effect and interactive ability , It has very promising future in applicable fields. Compression of multiview image data is one of the key techniques in multiview images processing, because the bandwidth necessary of storage and transmission linearly increases with the number of camera channels. This paper proposes several schemes to compress multiview images with view prediction, combining the characteristic of multiview images in 2D camera array, According to the differences in numbers or positions of I frames in one group of picture （GOP）, three types of multiview image coding structures and nine kinds of typical reference models are presented. Meanwhile, the rate-distortion is analyzed and compared. Experimental results show that the scheme with multilevel-B-frames is 0.2-0.8 dB larger than those of other schemes.

关 键 词：二维相机阵列 多视点图像压缩 编码结构 率失真性能 

分 类 号：TN911.73[电子电信—通信与信息系统]