Algorithm Research On Face Detection And Face Recognition

Since the face recognition had been proposed, it had always been paid somuch attention.From the cognitive perspective, the biomi metic patternrecognition method had overcome the samples partition problem that exited inthe traditional pattern recognition method. Though the study of a large numberof related literatures, we used the biomimetic pattern recognition theory,focusing on how to detect faces and extract the features of face sampleseffectively, as well as space covers and recognition in the high dimensionalfeature space,we have mainly researched on the following two aspects:(1)We have proposed an algorithm of face detection based on skin colormodel.First and foremost,we did the preprocessing job on face images,whichwill remove the noise and make the histogram of face images equalized.Then wetransformed the color space of face images from RGB to YCrCb and constructedthe skin color model.After that,we used the Bayes criterion to judge whether thepixels belong to the skin color or not,and then made the skin color areas intovalue 0,made other areas into value 1.Lastly,we conformed the precise positionof faces by using of information of face contour.Through the experiment results,verified that our method can adapt the change of both illumination and postures.(2) We have proposed an algorithm of face recognition based on biomimeticpattern recognition method.First and foremost,we introduced the principle ofDCT and LDA,and then we extracted the face features by using that twomethods,which reduced the dimension of face features.Secondly,we mapped theextracted features into high dimensional feature space through polynomialkernel function.Lastly,we used the theory of high dimensional space coverage torealize the classification of sample faces. By doing various experiments on theORL and Yale face database, the validity of our method has been proved.