| Image segmentation has been widely used in image processing, with many kinds of algorithms. In recent years, as new image segmentation way, pulse coupled neural network has gradually become a hot spot; this article does some research on metallographic image segmentation with complex background by PCNN.In this paper, an improved top-hat morphological enhancing algorithm has been proposed. In order to make the metallographic images for segmentation, we use this way to enhance the images by adjusting two new parameters.Then by learning the pulse coupled neural network model, some improvements are as follows. The parameters in this model are so many and complex that we need to simplify them and reduce the running time. Since there is a problem that it is hard to set the model's parameters, here we have do some research on setting parameters according to the existing methods. Segmentation has been the focus, and how to improve the segmentation is being people's research. Using this model to process metallographic images with complex background, will loss much edge information so that appears much error segmentations. In this paper, we combine SUSAN edge detection and PCNN model. Using edge matrix to initial PCNN threshold matrix, the neurons of image edge will be firing more suitable, so as to improve the accuracy.Finally, we propose an improved PCNN image segmentation algorithm based on morphological enhancing. The algorithm uses improved top-hat transform as a pretreatment before segmentation firstly. Then the improved PCNN model has been used to do image segmentation. We use mutual information function to judge the optimal segmented images. This paper combine an objective evaluation based on detection probability ratio and subjective evaluation of artificial way to analyze the experimental results. According to the experiments, using this algorithm to deal with the metallographic images with complex background will get more edge information than before. |
PDF Full Text Request