Karyote segmentation based on color-texture features

Karyote segmentation, a significant procedure to computer aided leukemia diagnosis, constitutes a difficult problem due to the morphological complexity, categorical diversity and distributional density of bone marrow cells. Most of the current techniques are based on the internal consistency of the karyote´s intensities with little or none considerations on color or texture properties. However, neither kernels nor cytoplasms exhibit consistent colors or structures. To secure better segmenting results, an approach based on color feature analysis together with LBP texture features, which considers both pixel level and regional level characteristics of karyotes, is proposed. Experiments are done and results show that our approach is simple and effective.