Investigation of low-level edge feature extraction using three blocks

Edge feature extraction is a necessary and important stage in edge detection. In order to extract edge features, all raw pixels in a local region are typically divided into two blocks based on a given direction. To reduce computational cost, a new method is proposed to split these pixels into three blocks so that the horizontal and vertical directions are considered at the same time. In this study, how to design three blocks to extract pixels for edge detection is investigated. The results show that using three blocks simultaneously to extract edge features reduces computational cost, but retains similar detection accuracy, compared with using two blocks based on the horizontal and vertical directions separately. Also, when one-way ANOVA F-tests are employed to filter noise, using three blocks once is better than using two blocks twice based on the horizontal and vertical directions to extract edge features from images under impulse noise.