Effect of Low-Pass Filters as a Shi-Tomasi Corner Detector's Window Functions

Abstract

The aim of this paper is to introduce an innovative way of using a low-pass spatial filter instead of window function of Shi- Tomasi corner detector, which is an enhancement of the Harris corner detector. The paper thus includes verification of the validity of this method on a reference image, and a comparison of different low-pass filters on test images. Particular spatial filters are linear smoothing filters mean filtering and Gaussian blur, and non-linear median filtering. The reason for this is to get an easy-to-implement algorithm for different architectures, such as a graphics card using shaders, which will allow fast processing of the input image even in real-time applications on less powerful devices, for example smartphones. This method can be applied not only to Shi-Tomasi corner detector, but also contains a formula for its use in the classic Harris corner detector. After a corresponding conversion of the evaluation function, the method can also be used for other variants of this corner detector, such as Noble detector.