Deformation tolerant generalized Hough transform for sketch-based image retrieval in complex scenes

摘要: Sketch-based image retrieval systems need to handle two main problems. First of all, they have recognize shapes similar but not necessarily identical the user's query. Hence, exact object identification techniques do fit in this case. The second problem is selection features compare with sketch. In domain-independent visual repositories, real-life images non-uniform background and possible occluding objects make task particularly hard. We address proposing a variant well-known Generalized Hough Transform (GHT), which robust technique for unsegmented images. Moreover, we solve first modifying GHT deal an inexact matching problem. paper, show how idea can be efficiently accurately realized. Experimental results are shown different databases real,