摘要
头肩序列图像是可视电话、会议电视等视频业务中常见的一种图像模式 .本文将SPIHT算法与人脸检测算法相结合 ,提出一种嵌入式的头肩图像编码方法 .首先采用人脸检测技术确定头肩图像中的人脸区域 ;然后产生人脸区的掩膜 ,并对LL子带中人脸区掩膜内的小波系数进行定标 ,以保证人脸区的小波系数能优先编码 ;最后采用SPIHT算法对人脸区和背景区进行编码 .实验结果表明 ,与原来的SPIHT方法相比 ,这种方法可以保证人脸区域的重建质量好于背景区的重建图像 .同时压缩后的码流仍然具有嵌入的特性 ,支持渐进传输 .
Human face detection algorithm is combined with the set partitioning in hierarchical trees (SPIHT) algorithm, and an embedded head-shoulder image coding method is proposed. The human face detection algorithm is employed firstly to determine human face region in a head-shoulder image. The human face region mask is generated and the wavelet coefficients in the human face region mask of LL subband is scaled to ensure that they are encoded with higher priority. Finally, the SPIHT algorithm is implanted to encode human face region and background region, The experimental results show that the reconstructed quality of human face region is much better than that of background region. At the same time, the encoded bitstream is fully embedded and supports progressive transmission.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2003年第12期1832-1834,共3页
Acta Electronica Sinica
基金
8 63计划(No.2 0 0 1AA1 2 1 0 61
2 0 0 2AA1230 4 1)
国家自然科学基金 (No .90 1 0 4 0 1 3
No .90 30 4 0 0 1)
