ACCURATE CELL SEGMENTATION IN BLOOD SMEAR IMAGES BASED ON COLOR ANALYSIS AND CNN MODELS

Abstract. Nowadays, automated blood cell evaluation play a major role in the classification and diagnosis of diseases. Despite many possible ways to segment cells, recognition efficiency remains insufficient, especially when different types overlap. Also, one should not forget about cells structure complexity. Image segmentation image are main stages this problem. At same time, smear images is considered most important stage disease detection systems. Often performed as separate mapping for white red cells. We propose another problem statement that uses capabilities supervised unsupervised CNNs semantic objects sizes shapes. CNN has encoder-decoder architecture builds pseudo-color map. tested several models using color spaces converting initial from RGB Lab, HSV CMYK obtained promising experimental results microscopic datasets such CellaVision DM96, All-IDB Blood Cell Detection.