Texture Segmentation using Modified Range filter with Adaptive PSO

Abstract

Texture segmentation is considered as the process of dividing an image into texture regions. Each region contains of similar group of pixels. Texture segmentation was made to chili x-ray images by the adoption of Gabor filter. The texture regions obtained from the Gabor filter are fed into the texture filter namely Range filter. The results of the Range filter are not able to understand the seeds and fungus affected regions are intertwined with each other in the chili images. In order to obtain clear differentiation some modifications were made in the Range filter, known as Modified Range filter. It works in the 5x5 neighborhood pixels instead of taking 3x3 neighborhood pixels. Modified Range filter selects the pixels from the horizontal, the vertical and the diagonal neighbor for computation of new range value by the proposed function. Adaptive Particle Swarm Optimization was applied to preserve the shape of the chili image. Mean absolute error was used as fitness function for the Adaptive PSO. The result obtained from Adaptive PSO were also contains clumsy of chili seeds. To rectify the above problem, a change in the inertia weight of Adaptive PSO was implemented. Now the results obtained are clear i.e. chili seeds are separated realistically. To ensure the obtained results of two performance measures were taken namely Uniformity and Gray level contrast. Average of the two measures was taken as final evaluation criteria. The results also infer that the Modified Range filter along with change in inertia weight of Adaptive PSO yields good segmentation result than the others. Key Words: Texture segmentation, Modified Range filter, Gabor filter, Adaptive PSO, Uniformity, Gray level contrast.