Research On Edge Detection Algorithm Employing Multiple Neighbor Cellular Neural Networks With Optimization Algorithm And Linear Matrix Inequality

Edge detection is an important part in the digital image processing.Some new edge detection algorithms are designed from the two aspectsof coupled cellular neural and uncoupled cellular neural networks, whichis based on the detail discussed in the classic edge detection algorithmsand cellular neural networks with theoretical models, cloning templatedesign methods and its applications. In the following, the main contentsare listed:1. An efficient edge detection algorithm using coupled cellularneural networks optimized by differential evolution is researched. Thisalgorithm uses completely stable template design method and also appliestemplates learning setup, in order to search for the ideal edge detectiontemplate, at the same time taking into account the influence of amongmultiple neighborhood cells. The simulation results show that theperformance of the proposed edge detector is superior to the classic edge detection algorithms and can correctly extract the edge, which is not bent;2. An edge detection algorithm exploring coupled cellular neuralnetworks with particale swarm optimization is researched. The simulationresults show that this algorithm has the following advantages: One is thatthe less number of particles are used; Two is that the less number ofiterations are taken; Three is that the effect of the detection is better thanthe classic edge detection algorithms; Four is that the effect of thedetection is more superious than the algorithm of based on the geneticalgorithm and clonal selection algorithm for edge detection. At the sametime, the simulation results also indicate that the cellular neural networkedge detector in performance depends on the efficiency of theoptimization algorithm.3. Image cancellation algorithm using coupled cellular neuralnetworks and linear matrix inequality is researched. Based on Lyapunovstability theorem, a criterion for global asymptotical stability of a uniqueequilibrium of the noise reduction cellular neural networks system, i.e. the sloving of feedback template is gave.Using the property of saturationnonlinearity of cellular neural networks, the inequality equations solvingincloduing controlled template and threshold is given. Peak signal tonoise ration is introduced into simulation experiment and the results showthat the algorithm not only can remove the noise effectively, but also doesnot destroy the edge information, which have the performance of a robustdenoising algorithm.4. Image edge detection algorithm using uncoupled cellular neuralnetworks and linear matrix inequality is researched. This algorithm hastwo aspects. When the center elements of feedback template a>1, thisalgorithm guarantees that the uncoupled cellular neural network systemcan output and extract ideal edge with finite iterations. When the centerelements of feedback template a equals to one, this algorithm can obtainstable edges by limited iterations, the simulation results show that thisalgorithm is superior to the classic edge detection algorithms, and at thesame can well obt ain the texture information of images, so this algoritm is very suitable forimage segmentation.