18 November 1999 Multichannel pattern recognition of color images
Author Affiliations +
Proceedings Volume 3904, Fourth International Conference on Correlation Optics; (1999) https://doi.org/10.1117/12.370405
Event: International Conference on Correlation Optics, 1999, Chernivsti, Ukraine
Abstract
In this paper we present the research line that we develop in pattern recognition of multichannel images, centered in the application of color images. Pattern recognition is performed through a multichannel correlation process. The correlation is applied to each channel, red, green and blue of the color image. The final recognition result is obtained by a combination of the information of the three monochromatic correlations. Two different approaches are proposed in order to improve the discrimination capability of the multichannel process. First, element-wise transformations over the multichannel images are used in order to enhance differences between channels. Then, the information in each channel is independent and the autocorrelation is enhanced with respect to the cross- correlations. The second approach involves the optimization of the matched phase-only filters used in every channel. This optimization is performed by means of a region of support. They are two complementary techniques that increase the discrimination capability and eliminate false alarms. The result is a better performance of the multichannel correlator for pattern recognition.
© (1999) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Maria Josefa Yzuel, Juan Campos, Ignacio Soriano Moreno, Esmail Ahouzi, Abdelaziz Attaleb, "Multichannel pattern recognition of color images", Proc. SPIE 3904, Fourth International Conference on Correlation Optics, (18 November 1999); doi: 10.1117/12.370405; https://doi.org/10.1117/12.370405
PROCEEDINGS
12 PAGES


SHARE
Back to Top