Getting poster data...
Vladyslav Luzin, Vladimir Ovcharenko, Anton Yeryomin (Lugansk state medical university, Lugansk, Ukraine)Aim is to compare the different binarization algorithms in histology. Methods. We carried out a comparative analysis of histological images binarization through a variety of automated methods in this paper: by Otsu, Niblack, Sauvola, Christian and Bernsan methods. Color photos of micro-histological preparations have certain characteristics that affect the final result of binarization. First, they are tend to have uniform illumination (at high magnification), high sharpness and contrast. Second, they have a poor color gamut and uniform background (usually 2-3 dye are used) depending on the methods of coloring. Third, they can have a large number of small objects, such as cell nuclei, and we need to keep its shape and size. Results. Christian method is the most successful method for binarization of the considered algorithms, but in some cases it has only a satisfactory rating. This suggests a need to further improve existing and develop new binarization algorithm, possibly with integration to achieve better results with modern means of recognition, adapted to the processing of histological photo.