Texture Segmentation of Urinary Sediment Image based on a Weighted Gaussian Mixture Model with Markov Random Fields.

In clinical laboratory studies, it is of importance for urinary sediment images because the composition and quantity of specific cells contained in the images revealing information for the diagnosis of urinary and renal diseases in humans. In this paper, we propose a method for the segmentation of urine sediment image using the magnification of 20-fold microscopy based on Markov model. This method selects sum average feature derived from the spatial gray co-occurrence matrix for the classification in the neighborhood window with the size of 7×7. First, each pixel in the image is evaluated by its intensity dispersion in comparison with the average intensity of the 7×7 neighborhood, and the pixels with higher dispersion would be further proceeded by Gaussian mixture model. Here the weights are designed feature distance matrix within the window to obtain similar, so as to re-estimate the features. Edge positioning is more accurate and clear compared with the traditional morphological methods by using the method of texture segmentation about urinary sediment image in this paper which can make it avoids cell adhesion caused by too closer proximity of cells and there is no glitch in the target area.