Abstract: The systematic approach that bridges the appearance gap for Sketch Based Image Retrieval System. The existence of noisy edge on photo realistic images is factor in enlargement of appearance gap and significantly degrades performance. To bridge this gap the system is proposed that consisting the framework of line segment based descriptor named as Histogram of Line Relationship (HLR) and a noise reduction algorithm as Boundary selection Algorithm. HLR Sketches and extracted edges of photo realistic images as series of Piece wise line segment and capture’s the relationship between them. Based on the HLR the object boundary selection algorithm reduces the impact of noisy edges by selecting the shapes edges .A fast method is applied to efficiently and the solution for object boundary selection algorithm. Multiple hypotheses are generated for descriptors by hypothetical edge selection. The selection algorithm is formulated to and the best combination of hypotheses to maximize the retrieval score. To reduce the distraction of false matches in the scoring process, two constraints on spatial and coherent aspects are introduced. By comparing the proposed HLR with state-of-the art descriptor it show that HLR descriptor outperforms them Combined with the object boundary selection algorithm, the framework significantly improves SBIRs performance.

Keywords: Large Scale Sketch Retrieval, Line Segment Based Descriptor, Histogram of Line Relationship (HLR), Object Boundary Selection Algorithm, etc.