Article ID Journal Published Year Pages File Type
4948194 Neurocomputing 2016 8 Pages PDF
Abstract

Generally, discrete orthogonal moments are difficult to induce rotation invariants. Based on relationship between Tchebichef polynomials and power series, we propose a new algorithm to compute rotation invariants of Tchebichef moments. The translation and scale invariants of Tchebichef moments are achieved by pre-normalizing the image to a standard image. Selected invariants of Tchebichef moments form a new effective shape feature for image retrieval. The retrieval performance of the proposed descriptor is compared with radial Tchebichef moment invariants and two kinds of Zernike moment invariants. Retrieval experiment results show that the proposed shape feature is robust to deformations generated by image shape rotation and scaling.

Related Topics
Physical Sciences and Engineering Computer Science Artificial Intelligence
Authors
, ,