Automatic recognition of white blood cells using weighted two phase test sample sparse representation

Microscopic images of blood are very important among the various medical images. One of the most important applications is to diagnosis blood disorders and its types like blood cancer. The main issue to diagnosis blood cancers is white blood globule either mature or not. There are many problem during using image processing to investigate white blood Cell can be mentioned as non-uniformity of colors, different brightness of images, variety of images, different size and texture of images, inherency of white cells in bone marrow images and adjoining of white cells to other blood parts like red blood cell. This paper used Gram-Schmidt orthogonalization process to obtain perpendicular bases that results in segmentation of blood cell kernels. To extract the Cytoplasm borders around the kernel, Variational Level Set Formulation of Active Contours Without Re-initialization method has been used. The main contribution of this paper is that after segmentation, the LBP has been extracted by converting colors in a new space YCBCR on the color factors of each channel so as to extract features. Afterwards by using WTPSSR classification approach and 10-fold valuation the precision of 95.56 has been obtained.