Abstract
We describe the circular-harmonic (CH) image CHmr, which is based on CH components for rotationally invariant pattern recognition. CH components of the order m, derived from an image in polar coordinates, are used to form a two-dimensional space together with the radial variable r. Filtering the CHmr image leads to a reference image with some rotational invariance. For a narrow-pass filter we have a single CH component with full rotation invariance; for an all-pass filter we have the original image with no rotational invariance; for a low-pass filter we form a reference image containing multiple circular harmonics with partial rotation invariance. Results of numerical simulations and optical experiments with a joint transform correlator are given that illustrate partial-rotation-invariant recognition for human face images.
© 1997 Optical Society of America
Full Article | PDF ArticleMore Like This
Pascuala Garcia-Martinez, Carlos Ferreira, Javier Garcia, and Henri H. Arsenault
Appl. Opt. 39(5) 776-781 (2000)
Oscar Gualdrón, Josep Nicolás, Juan Campos, and Maria J. Yzuel
Appl. Opt. 42(8) 1434-1440 (2003)
Shuqun Zhang and Mohammad A. Karim
Appl. Opt. 38(35) 7228-7237 (1999)