一种基于JPEG2000的人脸感兴趣区域编码新方法  

A New Algorithm for Coding of Region-of-Interest on face based on JPEG2000

作　　者：周登峰[1] 叶水生[1] 胡少华[1] 

机构地区：[1]南昌航空大学计算机学院,江西南昌330063

出　　处：《微计算机信息》2010年第23期200-202,共3页Control & Automation

摘　　要：JPEG2000的感兴趣区域编码较好的解决了压缩比与重构图像质量之间的矛盾,已经广泛地应用在了医学、无线传输等诸多领域,但以往的感兴趣区域都是由人为划定的,这大大降低了工作效率。针对这一问题,本文提出了一种基于AdaBoost和肤色检测的快速人脸检测算法,将人脸感兴趣区域的自动检测引入到了JPEG2000编码中。实验结果表明,与AdaBoost算法相比,该算法在不明显影响检测速度的基础上可以大大降低误检率。The Region-of-Interest coding in JPEG2000 is widely applied in many regions such as medical, wireless transmission etc, as it can solve the contradiction between the image quality and the compression rate, but the region of interest is mostly defined by mankind, and that decrease working efficiency a lot.To this point, a new algorithm for fast face detect based on AdaBoost and Skin color detection is proposed.The experiments results shows that the fault detect probability is reduced obviously compared to Ad-aBoost.

关 键 词：人脸检测 JPEG2000 感兴趣区域 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]