Abstract

Experimental data on the recognition of fragmented contour images with and without noise are compared with the results of recognition process modeling. A reliable approximation in the matched filtering model for erroneous responses in both stimulation cases was obtained only when the contours were replaced with the image silhouettes. The number of correct responses depended on the contour lengths for the case of images without noise and on the number of extended oriented contour sections for the case of noisy images. This indicates the significant role of the extraction of orientations related to the law of good continuation known from Gestalt psychology in the recognition of noisy images. Therefore, differences between noisy fragmented images and fragmented images without noise were determined in the modeling of the recognition process; i.e., the recognition dependence on the background or target environment was demonstrated.

© 2021 Optical Society of America

Optimal incoherent correlator for noisy gray-tone image recognition

Jianping Ding, Masahide Itoh, and Toyohiko Yatagai
Opt. Lett. 20(23) 2411-2413 (1995)

Improved Classification Robustness for Noisy Cell Images Represented as Principal-Component Projections in a Hybrid Recognition System

Maricor Soriano and Caesar Saloma
Appl. Opt. 37(17) 3628-3638 (1998)

The informational difference concept in analyzing target recognition issues

Dan Sheffer and Dov Ingman
J. Opt. Soc. Am. A 14(7) 1431-1438 (1997)

Different strategies in optical recognition of polychromatic images

María S. Millán, María J. Yzuel, Juan Campos, and Carlos Ferreira
Appl. Opt. 31(14) 2560-2567 (1992)

Hyperefficient detection of targets in noisy images

Ravindra M. Manjeshwar and David L. Wilson
J. Opt. Soc. Am. A 18(3) 507-513 (2001)