A Patch-based Sparse Representation for Sketch Recognition

Categorizing free-hand human sketches has profoundimplications in applications such as human computerinteraction and image retrieval. The task is non-trivial due tothe iconic nature of sketches, signified by large variances in bothappearance and structure when compared with photographs. Oneof the most fundamental problem is how to effectively describea sketch image. Many existing descriptors, such as Histogramof Oriented Gradients (HOG) and Shape Context (SC), haveachieved great success. Moreover, some works have attemptedto design features specifically engineered for sketches, such asSymmetric-aware Flip Invariant Sketch Histogram (SYM-FISH).We present a novel patch-based sparse representation (PSR)for describing sketch image and it is evaluated under a sketchrecognition framework. Extensive experiments on a large scalehuman drawn sketch dataset demonstrate the effectiveness of theproposed method