Advanced K-views Algorithms

Abstract

This chapter introduces the weighted K-views voting algorithm (K-views-V) and its fast version called the fast K-views-V algorithm. These methods are developed to improve K-views template (K-views-T) and K-views datagram (K-views-D) algorithms for image texture classification. The fast K-views-V algorithm uses a voting method for texture classification and an accelerating method based on the efficient summed square image (SSI) scheme as well as the fast Fourier transform (FFT) to enable overall faster processing while the K-views-V only uses the voting method. In classifying a pixel to a texture class in the K-views-V algorithm, it will be based on the weighted voting method among the “promising” members in the neighborhood of a pixel being classified. In other words, this neighborhood consists of all the views, and each view has this pixel in its territory. Experimental results on some textural images show that this K-views-V algorithm gives higher classification accuracy than the K-views-T and K-views-D algorithms, and improves the accurate classification of pixels near the boundary between textures. In addition, the acceleration method improves the processing speed of the K-views-V algorithm. Compared with the results from earlier K-views algorithms and those of the gray-level co-occurrence matrix (GLCM), the K-views-V algorithm is more robust, fast, and accurate. A comparison on the classified results with the selection of parameters on the view size, sub-image size, and the number of characteristic views is provided in this chapter.

Kindness in words creates confidence. Kindness in thinking creates profoundness. Kindness in giving creates love.

—Lao Tzu