An improved SAM algorithm for red blood cells and white blood cells segmentation

Abstract

The segmentation of red blood cells and white blood cells has important research value in the field of rheological properties of blood and the pathogenesis of some diseases. And it is the reflection of bone hematopoietic state, blood diseases and other diseases. Especially for the diagnosis of blood diseases, the detection and prevention of treatment process, there is high value of clinical research. The separation of red blood cells and white blood cells using hyperspectral remote sensing image processing is a new field that it is essentially different from traditional multi spectral classification. Because of the different chemical composition and molecular space structure of red blood cells and white blood cells, it results in different spectrum. Each pixel of hyperspectral image can obtain a unique continuous spectral curve, and it can be compared with the spectral curves which are known to obtain target object. So the author designs a new analytical method which is based on the various processing methods of hyperspectral image. First of all, using the BandMax wizard to lock target image and band based on target detection; secondly, conducting differential search algorithm based on the blind signal; thirdly, using an improved algorithm—based on SAM combined with SID algorithm; finally, using advanced filtering method to get clearer image information. In this paper, it focuses on the effective extraction and improves the classification accuracy of white blood cells.