A Kernel Fuzzy Clustering Infrared Image Segmentation Algorithm Based on Histogram and Spatial Restraint

Because the contrast of the image for guiding the highspeed infrared air-to-air missile is low, its signal to noise ratio is poor and the target and its background gray-scale coupling is strong, the paper analyzes the reasons why the threshold value segmentation method and the fuzzy C-means clustering method have the over-segmentation and under-segmentation in segmenting the above type of image. Hence we propose the kernel fuzzy clustering segmentation algorithm based on histogram and spatial constraint, which utilizes the global first-moment histogram of the infrared image to restrict the number of clusters and the clustering center, improves the spatial correlation function that fully manifests the correlations among pixels inside a neighbor domain and reconstructs the membership degree matrix and the clustering central function, thus segmenting the infrared image with the kernel fuzzy clustering algorithm. The results on the experiments on a sequential infrared image show preliminarily that, compared with the traditional threshold value segmentation algorithm, the fuzzy C-means segmentation algorithm and the kernel fuzzy clustering algorithm, the improved algorithm proposed in the paper can reduce entropy segmentation by about 60% on average and increase the correlation degrees among clusters by around 10%, thus enhancing to a certain extent the efficiency and precision for segmenting the fuzzy image whose target gray-scale and background gray-scale are strongly coupled.