Application And Research On Pulse Coupled Neural Network In Face Retrieval

Artificial Neural Network (ANN) is an inter-discipline composed of manydifferent fields, which is to imitate the human brain cognitive processes. It iscomplicatedly structured like a network being composed of a lot of artificial neurons.Artificial neuron of different Network has its own function and learning algorithm,thus appearing all kinds of artificial neural network model, which is in order to solveproblems that exist in practical engineering or science.It through more than half acentury, that from Perceptron to Pulse Coupled Neural Network(PCNN). In morethan20years after the study, the pulse coupled neural network is gradually usedwidely in the field of image processing.Compared with the traditional neural network, PCNN does not need to study ortrain, but has the ability of self learning and unsupervised, and also has a similargroup of neurons distributed synchronization characteristics of the pulse. PCNN hassome special advantages when it is used in digital image processing: this model cangroup the pixels, which are similar at two dimensional apace or gray level, and reducethe local gray difference of the image, make up for the local small gap at the sametime. The output of binary image retains some information of input image such astexture, shape, edge and so on. It is invariant to input image scale, translation androtation to some extent, also has a certain resistance to noise at the same time.Because of these features, PCNN is widely used in various fields of image processing,such as image segmentation, image enhancement, image edge information detectionand so on.For the human facial feature exaction, this paper use standard PCNN model toextract facial feature on the basis of analysis the superiority of standard PCNN modelwhen it is used for image feature extraction, then analysis the deficiency of featureextraction of similar face and different face.In order to solve these shortcomings,improve the standard PCNN model, namely the PCNN-X model, and proved thatPCNN-X model in face feature extraction is effective through the experiment. For the face retrieval, in order to prove that the PCNN-X model is applicable toface library retrieval, Face when needs a lot of face image retrieval, this paper probesinto the method for extracting facial color character image, the character imagepreprocessing, face skin color model to extract the face image is established. this paperuse the standard ORL face database to search face. During the face retrieval, theentropy sequence, which is extracted by PCNN-X is transformed to be binarysequences by the use of Random Hyperplane Hash(RHH), then the binary sequencesdatabase of face image is established, finally the similarity between images ismeasured by hamming distance, make the face retrieval come true.