Abstract— Diffusion geometry plays a vital role in shape analysis and object recognition. It evokes from propagation of heat on the object surface. This derives the intrinsic or invariant features of the surface. The heat kernel signature (HKS) based on heat diffusion suffers with the problem of scale sensitivity. This is resolved by the scale invariant heat kernel signature (SIHSK). It involves the Fourier descriptors for providing scale invariance property. This method considers only orthogonal features. The present paper has extended this by considering the curvature properties to SIHSK viz., curvature heat diffusion method with SIHSK (CSIHSK). The proposed method adopted the Modified Euclidean distance for shape similarity measure and is experimented over three standard databases. The results prove the efficiency of the proposed method than that reported for other descriptor of concurrent interests.
Index Terms— LOB operator, eigen modes, eigen values, heat kernel, fourier transform.
M. Radhika Mani was with Pragati Enineering College, Surampalem, A.P., India (e-mail: radhika_madireddy@yahoo.com).
G. P. S. Varma is with SRKR Engineering College, Bhimavaram, A.P., India (e-mail: gpsvarma@yahoo.com). Potukuchidm and C. Satyanarayana are with JNT University Kakinada, Kakinada, A.P., India (e-mail: potukuchidm@yahoo.com, chsatyanarayana@yahoo.com).
[PDF]
Cite: M. Radhika Mani, G. P. S. Varma, Potukuchidm, and C. Satyanarayana, " A Novel Approach of Curvature Based Heat Diffusion for Shape Based Object Recognition," International Journal of Information and Electronics Engineering vol. 5, no. 4, pp. 265-269, 2015.
