| Summary: | 博士 === 國防大學理工學院 === 國防科學研究所 === 107 === Object contour is a key feature to describe the object. Therefore, how to effectively detect the contour of an object and complete the object segmentation to achieve object recognition processing is an extremely challenging problem. The edges are the basic element to construct the contour of an object, which contain most of the information in an image. Hence the edge detection method can be regarded as the cornerstone of image processing. Good edge detection results can be used to further extract the required information through the rich edge texture information, and to achieve object detection, object segmentation and recognition applications. In order to detect rich texture, this study proposes an edge detection algorithm. The algorithm uses the extension of the texture change between blocks to construct the edges of an object. Therefore, the longer edge texture in an image will be retained as an effective edge texture feature. Compared with the traditional edge detection methods, the threshold value is used to preserve the edges of an object, and the proposed algorithm can effectively reduce the edge loss caused by the unsuitable threshold value setting. Finally, through the verification of the experimental results, the algorithm proposes in this paper can not only overcome the problem caused by the unsuitable threshold setting, but also can detect the rich object edge information.
|
|---|
