Methods based on local image features have recently shown promise for textureclassification tasks, especially in the presence of large intra-class variationdue to illumination, scale, and viewpoint changes. Inspired by the theories ofimage structure analysis, this paper presents a simple, efficient, yet robustdescriptor namely local jet pattern (LJP) for texture classification. In thisapproach, a jet space representation of a texture image is derived from a setof derivatives of Gaussian (DtGs) filter responses up to second order, socalled local jet vectors (LJV), which also satisfy the Scale Space properties.The LJP is obtained by utilizing the relationship of center pixel with thelocal neighborhood information in jet space. Finally, the feature vector of atexture region is formed by concatenating the histogram of LJP for all elementsof LJV. All DtGs responses up to second order together preserves the intrinsiclocal image structure, and achieves invariance to scale, rotation, andreflection. This allows us to develop a texture classification framework whichis discriminative and robust. Extensive experiments on five standard textureimage databases, employing nearest subspace classifier (NSC), the proposeddescriptor achieves 100%, 99.92%, 99.75%, 99.16%, and 99.65% accuracy forOutex_TC-00010 (Outex_TC10), and Outex_TC-00012 (Outex_TC12), KTH-TIPS,Brodatz, CUReT, respectively, which are outperforms the state-of-the-artmethods.
展开▼