Pattern Recognition based on Single-Channel Correlation of Color Images

Different methods of color image correlation are proposed. They are based on a "binary phase only filter" (BPOF) and the segmentation method. The objective of merging simultaneously "single-channel correlations" is to save computing time, to alleviate complexity and to reduce cost. These methods are investigated to determine the better match adaptive one to our applications. Multi-channel processing is applied to color images by decomposing the image into its three main color components "red, green and blue " (RGB). To each color components, one channel is assigned and processed separately. Then the respective outputs are combined. Our new contribution consists in putting together the respective spectra of the color channels to obtain a single spectrum. Instead of three spectral correlations, we perform only a grouped color correlation. After applying an optical Fourier transform (FT) on each channel, the three respective spectra are combined together. The correlation is carried out in the spectral domain and results in a compound spectrum. Indeed, relevant information should be selected from each spectrum of the three color components to form a unique spectrum which presents the core of the correlation. This resulting spectrum is then multiplied by the binary phase only filter (BPOF). Since a single phase filter is used, a significant reduction of the data base volume is achieved.