Ramanujan sums based image kernels for computer vision

摘要

In the recent history, kernel methods had established themselves as powerful tools for computer vision. In this paper we introduce an integer image kernel function based on Ramanujan Sums which finds its place in image vision. The paper proves the validity of kernel function theoretically and also shows the application of the kernel in image vision. Ramanujan Sums are based on number theory and hence the new kernel matrix will contain only the integer values. Since the image processing involves complex matrix manipulations, the processing based on the new kernel will be computationally effective. The paper shows the applicability of the kernel in various context of image processing. By applying the theory of Ramanujan Sums for image kernel, we will show the intervention of numerical mathematics in machine learning which gives new directions for future research.