An improved histogram of edge local orientations for sketch-based image retrieval

摘要: Content-based image retrieval requires a natural (e.g, photo) as query, but the absence of such query is usually reason for search. An easy way to express user using line-based hand-drawing, sketch, leading sketch-based retrieval. Few authors have addressed based on sketch and current approaches still keep low performance under scale, translation, rotation transformations. In this paper, we describe method computing efficiently histogram edge local orientations that call HELO. Our strategy applied in context fingerprint processing. This descriptor invariant scale translation To tackle problem, apply two normalization processes, one principal component analysis other polar coordinates. Finally, linearly combine distance measures. results show HELO significantly increases effectiveness comparison with state art.