Automatic design of color w-operators to segment white blood cells in images of acute lymphoblastic leukemia
This paper proposes the automatic design of color W-operators to segment white blood cells of acute lymphoblastic leukemia in images in RGB color space. The design of the proposed W-operators consists of two stages, one for training and one for testing. The training stage is carried out marginally for each of the RGB color space channels. For the training stage, we use different sets of n-pairs of images to obtain an array of observations, for each channel, with their respective labels observed in a window of size k. In the test stage, the three associated W-operators obtained in the previous stage are combined, obtaining a color W-operator. The proposed method was used to segment white blood cells with satisfactory results To validate this method, a free access database of Leukocyte Images for Segmentation and Classification (LISC) was used. From the results, it can be concluded that the proposed W-operator color allows better segmentation than those defined from a single channel or the linear combination of the three channels.