Polychromatic image fusion algorithm and fusion metric for automatized microscopes

Abstract

Abstract. We propose a new algorithm to determine the multifocus image fusion from several polychromatic images captured from the best focusing region where the best in focus image is included from a biological sample. This focusing region is built by including several images up and down starting from the Z position of the best image in focus. These captured RGB images are converted to YCbCr color space to have the color CbCr and intensity Y channels separated with the objective to preserve the color information of the best in focus image. Several approaches have been developed to fuse images, like those algorithms based on the wavelets transform, Laplacian, ratio, contrast or morphological pyramids selection, fusion by averaging, Bayesian methods, fuzzy sets, and artificial networks. However, this algorithm utilizes the Fourier approach by using the Y channel frequency content via analyzing the Fourier coefficients to retrieve the high frequencies to obtain the best possible characteristics of every captured image. After the completion of this process, we continue to construct the fused image with these coefficients and color information for the optimum in focus image in the YCbCr color space; as a result, we obtain a precise final RGB fused image.