Research On Edge Detection Of Color Images Based On Mathematical Morphology

Edge detection is one of the fundamental tasks in image processing andcomputer vision, because of its wide use in several techniques such as segmentation,object recognition, tracking, stereo analysis, data hiding and image coding. Edge isthe mutational part of gray value or structure information in image processing whichis one of the most basic features when distinguishing different objects. Color imagescontain more information than gray images, studies show that about90%edgeinformation of one color image is the same with its gray image, but the rest of10%cannot be detected in the gray image. Therefore, the study of color image edgeinformation is one of the frontier subjects in recent years.In the past, mathematical morphology was usually used to deal with two valueimages or gray images. But now, it is an ideal method in the filed of digital imageprocessing and computer vision, and its application in image processing has become anew research direction. Although there are some studies using mathematicalmorphology for color images, the application is also not much.Based on the study of two value morphology and gray scale morphology, andconsidering the disadvantage of edge detection in RGB color space, this paperintroduces the mathematical directly into the HSI color space. Firstly, we provide therelated definitions of the color alternating sequence filter operators of opening-closing(ASFOC c) and closing-opening (ASFCO c) based on multi-scale and multi-structuralelements in HSI color space. In order to avoid the shortcomings of the two filteroperators ofASFOC candASFCO c, we propose an improved color morphologicalhybrid filter (HF) by using the weighted mixed thought to combine the two filteringmethods simultaneously. And experiments show the proposed filter can not only filterout the noise effectively but also well protect the corner details. Then in view of theresearch of the color morphological hybrid filter (HF), this paper proposes a colorimage edge detection method based on color morphology multi-structural andmulti-scale and Omni directional elements in HSI color space with the thought ofmulti-structural and multi-scale and Omni directional elements and the knowledge ofinformation Entropy. Experiments show that the proposed algorithm has good noise resistance and retains more image characteristics, and it lays a solid foundation tostudy other related properties of the color image.