Face Region Of Interest Encoding Based On Adaboost Algorithm

JPEG2000, the new ISO/ITU-T Standard for still image encoding, had been finished at the end of 2000. Compared with the traditional JPEG image compression standard, it has lots of new features and functionalities, in which“Region of Interest”(ROI) Coding is undoubted a key point. After being proposed, ROI Coding had become a hot research point in still image compression area. According to people’s need and preference, they often have great interest in some particular part of the image. This is the ROI concept which can be used to make a balance between the compression ratio and the image reconstructing quality.The paper introduces the face detecting and locating method into JPEG2000 image compression system, which forms new still image compression architecture. Based on the new architecture, it implements automatic face ROI encoding software system in C++ language on Visual Studio 2005. The paper first discusses all stages in JPEG2000 standard thoroughly, including pre-pressing, wavelet transform, quantization, EBCOT algorithm and MQ encoder. Then it focuses on the EBCOT algorithm and ROI coding principle, gives implementation on MAXSHIFT ROI coding method and proposes fast ROI mask generating algorithm. Meanwhile, it gives a detailed analysis on face detecting and locating based on Adaboost algorithm, implements face detecting and locating module using Opencv Library. Finally, it introduces the face detecting and locating module into JPEG 2000, implements automatic face ROI encoding software system and verifies it vadility and feasibility through variety of experiments.