An region of interesting compression algorithm of still image is introduced. The algorithm that is based on embedded block coding with optimized truncation (EBCOT) encodes the interested region of the image. According to the character of image edge, an improved canny edge detection algorithm is proposed before the wavelet transform, without the participation of automatic extraction of artificial region-of-interest (ROI) by using the dilation and erosion operation in morphology. The ROI coding algorithms is analysed. ROI information can be encoded with high priority at the same time by con structing a weighted function, giving reasonable weight for ROI code block, and reducing the wavelet coefficients effect on the context region of the ROI code block. The proposed method improves the quality of the reconstructed image of ROI. Experiments show that the reconstructed ROI image quality is significantly improved under low bit rate or high, and the reconstructed image background region (BG) quality can be also improved.
Index Terms: ROI, Region Of interest, Embedded Block Coding

Automatic face identification of characters in movies has strained important do research benefit and lead to several attractive applications. It is a difficult problem due to the massive difference in the form of each character. Even though presented methods make obvious hopeful outcome in sparkling atmosphere, the performances are some degree of in compound movie scenes due to the noises generated at some stage in the face tracking and face clustering development. In this paper we present two schemes of global face-name matching based structure for robust character identification. The offerings of this work include the following.
1) A noise thoughtless character connection demonstration is integrated.
2) In introducing correct procedure based graph matching algorithm.
3) Difficult character changes are handled by all together graph screen and graph matching.
4) Further than presented character identification approaches, and then perform an in-depth understanding psychoanalysis by introducing two types of virtual noises. The proposed schemes display high-tech presentation on movie character identification in different genres of movies.
Keywords: Eigen Object Recognizer, video, face detect and recognize, training name, training face, character identification.