Segmentation of Microscopic Medical Images Using Local Binary Patterns Method

Abstract

This paper deals with the problems of segmentation of specific microscopic medical images for diagnostic purposes. The monitored structures are physiologically motile with right frequency and synchronization which is evaluated by analysis of high-speed video sequences containing microscopic samples. It means that the diagnostic of potential pathologies leads to analysis of particular images of the video sequence. Proper recognition of image contents and its division into objects of interests, artifacts and background is a task of image segmentation. There exist a vast range of segmentation methods in the area of image processing, however, the particularity of studied microscopic images and the requirements for precision mean significant difficulties and this opens the space for further research in this field. This paper proposes better and more precise segmentation of images containing respiratory epithelium, using the method of local binary patterns. Such technique focuses on local texture features of an image, which enables the reliable identification of even small and usually not contrast structures -- cilia. The great potential of this approach is in the possibility to identify the static cilia, which could not be recognized by standard frequency analysis of video sequences (due to the fact that they are actually not moving), and so improve the diagnostic of even more serious ciliary pathologies.