Reseach And Implementation Of The Region Of Interest(roi) Encoding Based JPEG2000

With the increasing development of multimedia and network technology, it is more important to improve compression ratio while meeting the different needs of users. ROI coding can largly meet such ruquirments. According to users’ preferences, an image can be divided into the background region(BG) and the region of interest(ROI), and then we use different encoding mechanism for different regions. After ROI coding, ROI is priority over BG, so ROI can be first restored and its reconstruction quality is better than BG. This feature is particularly important at low bit rates. JPEG2000 is a new standard of static image compression, it not only has more outstanding performance in compression than other image compression standard, but also has many new outstanding features, and one of these features is ROI encoding.In this thesis, we study the ROI coding technology based on JPEG2000 standard and complete its implementation. Combining with the use of DWT in the JPEG2000, we propose a method which is based on image’s resolution level property after DWT. In addition, we add the ROI extrating in this thesis, so the encoding system can automatically acquire the ROI befor ROI coding, improving system’s integrity.Firstly, we study and analysis some commonly used ROI enconding algorithms, including MaxShift, PSBShift and BbBShift, particularly the first two algorithms. MaxShift is easy to implement but it can not adjust the quality contrast of ROI and BG. PSBShift is better than MaxShift in adjusting the quality contrast of ROI and BG, but its adjustment sometimes still can’t meet the practical needs. In order to improve the shotcomings in these methods, combining JPEG2000 and the different property of these methods, we present a new method of ROI coding based on resolution levels after DWT. In this new method, the main idea is to contribute the expense of ROI DWT coefficients to compensate for BG. In other words, we treat some ROI DWT coefficients as BG DWT coefficient, so the quality of BG is improved in ths way. After applying this new method to MaxShift and PSBShift, the simulation results show that at the same bit rate, BG’s PSNR value can be increased 2dB~4dB while the quality of ROI is not apparently changed.In order to improve the integrity of the coding system in this thesis, we introduce the region of interest extraction for the image which is to be ROI encoded in the implementation. Face is treated as ROI in this thesis, so we choose the face detection as ROI extraction. Combining the application background and ROI extraction is ROI coding preprocessing stage, we choose skin color based face detection in this thesis.Finally, we combine the ROI coding and human face detection and complete an automatically ROI detecting and coding system.