Research On Morphological Associative Memories Based On Dynamic Kernel And Its Application

Associative memory is one of important capabilities of human brain cell.In recent years, morphological bidirectional associative memories (MBAM)[1,2] presented by G.X.Ritter is a latest advanced in this aspect. It overcomes disadvantage of the traditional associative memory model which has less storage capabilities and need many iterations. MBAM circumvents these disadvantages and obtains the success in processing binary image with pure noise but not gray-scale images. Recently, there are not many associative memory networks applied for gray-scale images. Number of associative memory networks applied for gray-scale images with random noise is few.Based on MBAM, morphological associative memories of dynamic kernel is presented in this paper. In this method, which is added with dynamic kernel, can be applied flexibly to recognize the image with random noise. Because the kernel of per original picture is not single, the results are decided by the recognized picture with the noise. The kernel is different with different recognized picture, so that we call it dynamic kernel. The method is better than the former, especially with more noise,more image data, which is verified by the experiments.Meanwhile, MBAM just be applied to binary images but not gray-scale images. In this paper, the method of dynamic kernel is also applied to gray-scale images, and achieves ideal results. It not only can revert the pictures with noise, but also can recognize the pictures, which is verified by the experiments...