Shen-Castan Based Edge Detection Methods for Bayer CFA Images

Color Filter Array (CFA) represents a mosaic of incomplete color information from a digital image. This paper presents two edge detection methods performing directly on CFA images, without the necessity of the demosaicing process, thus saving significant computation steps. First, existing methods for CFA images based on well-known Deriche recursive filters are revisited. Then, new algorithms based on Shen-Castan filters design are proposed. They correspond to recursive first-order filters, outperforming the complexity of other edge detection techniques. Finally, quantitative assessments based on synthesized images using normalized Figure of Merit evaluate the performances of the edge detection methods, while qualitative results based on real images are also reported, illustrating the new methods reliability.