Research Of Face Detection Based On Dsp

In this thesis, we summarize the technical characteristics of Digital Signal Processor (DSP) TMS320 DM642, and analyze its critical nature of both image and network's constitutional unit.We show that it is appropriate to choose DM642 as core processor when constructing a video-image processing system. Introducing platform development environments, such as CCS and DSP/BIOS, which are development software and embedded operating system respectively, is also part of our works.We review the research background and application of face detection within giving an analysis on its history. We also introduce some principles and basic techniques of Digital Image Processing (DIP). Then a new approach of face image edge-detection based on DSP is presented. As shown in the paper, this approach is considered both real-time speed processing and degree of edge's accuracy. We emphasize that our approach involving with a DSP-based rapid Face Detection algorithm, is from a practical perspective. We first use the method of maximal color-balance to adjust the color cast of images, and do complexional segmentation by combining the double color-space models. Improving the image's qualities, which achieved through mathematical morphological operation, is in our subsequent consideration. A confirmation of prior knowledge of human faces is also carried out in order to reduce the ratio of false detection. As a result, we are able to detect and locate the facial organs, such as eyes,nose,mouth within a rectangular box including face that obtained from previous steps.Paper divides the framework of system into four parts:capture and input of video image,detecting and handing,video display and network transmission, after of which is a introduction of the work processes of system. The RF-5 (Reference Framework 5) of design with high expansibility, as well as C and Assembler mixed language, are used to carry out the design and implementation of program of system. Then, system optimization has been done in aspects of C and assembler through CCS considering the Face Detection and characteristic of chip of DM642.In the end, related image results has been displayed and analyzed.