Modified Hough Transform for Images Containing Many Textured Regions
Images which have a lot of textured regions make the result of Hough transform (HT) very poor. This paper presents an improved HT that deals with such a textured image by diminishing the effect of noise edges and using weighted voting score. The method first eliminates the noise edges resulted from textured regions; then, the method casts votes for edges upon the accumulator array with weight score in accordance with the number of sequential votes. Our modified HT is efficient in that it produces important lines first such as verge of building, avoiding improper lines taken from the noise edges.
KeywordsEdge Point Edge Image Texture Region Canny Edge Detector Small Scale Parameter
Unable to display preview. Download preview PDF.
- 1.Schmid, C., Zisserman, A.: Automatic line matching across views. In: Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp. 666–671 (1997)Google Scholar
- 10.Hough, P.V.C.: Method and means for recognizing complex patterns, U.S. Patent No. 3069654 (1962)Google Scholar
- 12.Yuen, S.Y.K.: An approach to quantization of Hough space. In: Proc. 7th Scandinavian Conf. Image Analysis, pp. 733–740 (1991)Google Scholar