A Research On Image Processing And Retrieval Identification Technology Based On Intelligent Computing And PCNN

Along with the rapid development of multimedia technology and the World WideWeb, It’s an urgent problem for people to retrieve the image information from massiveimage data. Unlike the traditional text-based image retrieval method, CBIR become a newtechnology of large-scale multimedia database and efficient management and retrieval. Thecontent-based image retrieval built on computer vision, image understanding and imageanalysis involves many aspects of information retrieval, image processing, patternrecognition, computer graphics and database management technology. It becomes is a hotresearch topic in image information retrieval recently.This paper extracts image features on the basis of PCNN (Pulse Coupled NeuralNetworks) and applies them to image content retrieval. It studies intelligent computing andthe application of PCNN in image denoising, segmentation and recognition retrieval so asto form a research program of image processing and retrieval identification based onintelligent computing and PCNN.The results the paper has done can be summarized into the following five parts:1. Intelligent computing is applied to image segmentation. This paper summarizes thebasic theory and common methods of image segmentation, detailing the otsu andmaximum entropy theory and undertaking image segmentation combined with intelligentalgorithms. In this paper the classic genetic algorithm is selected in image segmentation tocombine itself respectively with otsu and maximum entropy theory and complete thecorresponding simulation experiments. It has also made a comparison with the simulationresults of otsu and graythresh that goes with matlab.2. Intelligent computing is applied to image recognition. In this paper BP neuralnetwork of intelligent computing is selected in image recognition so it analyses the designof BP neural classifier, input layer, hidden layer and output layer. It selects the traditionalgradient descent algorithm mainly to identify images containing numeric images andcompletes the simulation.3. PCNN together with the algorism maintaining the edge template is used in imagedenoising. Here it mainly processes the impulsive noise within the image, undertakes thesimulation and extracts the image edge after denoisng. The simulation experiment includesdenoising of the single static images containing noise and also that of the sequent dynamictarget images containing noise.4. The optimal PCNN parameter in immune algorithm is used in image segmentation and feature extraction. The paper combines the optimal PCNN parameter with maximumentropy and otsu respectively and completes the simulation. On this basis it calculates the“bright” points of the binary image each time the image is segmented so as to get the timesequence (time signature) as the image features. The research shows the time signature isinvariant when we conduct the image target’s rotation, translation, scale change, distortion.5. This paper forms a total solution for image recognition consisting of denoising,segmentation and feature extraction. It combines PCNN with maintaining the edgetemplate to undertake image denoising; It also combines optical PCNN parameter withmaximum entropy and otsu in image segmentation and applies time signature extraction inimage retrieval so as to finally complete the prototype system of content-based imageretrieval.