Research on the dual kernel based on the unified framework of Morphological Associative Memories

The Morphological Associative Memories (MAM) are a class of extremely new artificial neural networks. Although MAM has many advantages, it can not have a good performance of tolerating random noises and mixture noises. In order to solve this problem, some people have presented the kernel method, but there is still room to research. So we research on the kernel from another angle. On the one hand we present the concept of the dual kernel based on the unified Framework of Morphological Associative Memories, on the other hand, we research on its properties in-depth. At last, through the computing and simulation experiments of true-color images, we confirm that the dual kernel has the validity of tolerating random noise and mixture noise.