Parameter-Free Estimation of the Number of Texture Segmentations for Application of Satellite Imagery in Agriculture

A number of papers and articles have been published regarding the number of textures. However, all these approached adopt an exhaustive search for the optimal solution, and are not appropriate for real time application such as segmentation. Although our previous approach regarding the number of cluster was discovered, but it was hard to interpret a number of peaks in triple co-occurrence matrix as the number of textures. The best of our previous approach need to automatically find and interpret a number of peaks in triple co-occurrence matrix. In this research, the paper presents a simple, parameter-free method to estimate the number of textures for satellite imagery in agriculture using an image processing approach based on the triple co-occurrence matrix technique. A subsampling technique, smoothing filter and maxima function are proposed for reducing matrix size and time in order to count a number of peaks in triple co-occurrence matrix as the number of textures. The proposed method was tested using data from known the number of textures with satellite imagery. The result from the test confirms the effectiveness of the proposed method in finding the estimation the number of textures and compared the result with ground truth data.