Edge Detection in Satellite Images using Quaternion Convolution in Frequency Domain

摘要

Edge detection has been an interesting subject since the beginning of image processing. Lots of edge detection techniques on gray-scale images have been proposed in the last decades. In this paper, we present an edge detection approach in the frequency domain based on quaternion valued colour edge detecting filters and convolution with quaternion valued masks known as hypercomplex convolution or quaternion convolution. These filters, when applied to a color image, produce an almost grayscale image with colour edges where the original image had a sharp change of colour. These filters depend on quaternion rotation. The hypercomplex convolution is implemented in frequency domain using quaternion Fourier transforms. This procedure is suitable for color images and is tested on the Landsat colour images.