面向头肩序列图像的质量可精细扩展视频编码方法

FGS编码方法具有细粒度的可扩展能力,能很好地适应网络带宽的动态变化,被认为是一种适合于网络视频传输的编码方案.但现有的MPEG-4 FGS编码标准效率低,限制了其进一步的推广应用.因此,本文面向视频应用中常见的头肩序列图像,实现了一种质量可精细扩展的视频编码方法.该方法采用H.26L对基本层进行编码,采用基于DCT变换的SPIHT方法对原始图像与基本层重建图像之间的残差进行编码得到增强层的码流.然后将复杂背景下的人脸检测与跟踪技术与选择性增强技术结合起来,对人脸区域优先编码.实验结果表明,该方法不仅编码效率高于现有的MPEG-4 FGS标准,码流具有可精细扩展的特性,还可以选择性地提高人脸区域重建图像的主观感受水平.

Quality Fine Granular Scalability Video Coding Method for Head Shoulder Sequence Images

Fine granular scalability (FGS) coding method has fine-grained scalable capability and can be adapted to dynamic variation of network bandwidth.So it is considered as a good coding scheme suitable for video transmission over network.But current MPEG-4 FGS coding standard is not efficient,which will restrict its application.Therefore,a quality fine granular scalable coding scheme for head shoulder sequence images,which are commonly seen in video applications,is implemented in this paper.The base layer is encoded with H.26L and the residual signal between the original image and the reconstructed image from base layer is encoded with a DCT-based SPIHT coding method to achieve the enhancement layer bit stream.Automatic human face detection and tracing algorithm in a complex background is combined with selective enhancement technique to encode human face region with high priority.Experimental results show that the overall coding efficiency gain of this method is higher than that of MPEG-4 FGS standard and the bit stream is fine granular scalable.The subjective perceptual quality of reconstructed human face region can be selectively improved.