Identification of geometrical properties of red blood cells from light scattering images using SVD

In this paper an automatic identification method of geometrical properties of Red Blood Cells (RBCs) using light scattering images, is presented. A small number of features are estimated by the pixels' intensity projection into an RBC-space. The basis of the RBC-space is derived using the singular vectors of a set of known RBCs. The nearest neighbor rule is used to classify any image projection to the known RBC coordinates. Since, the dimension ability of the RBC-space is significantly lower than the whole scattering image-space, it is easier to compare projections than original images. Considering the above idea, a Singular Value Decomposition (SVD) approach is implemented in this work. The database includes 1188 simulated scattering images, obtained by means of the Boundary Element Method (BEM). The identification accuracy of the actual RBC shape is estimated using three feature sets in the presence of additive white Gaussian noise from 60 to 10 dB SNR, giving a mean error rate less than 1 percent of the actual RBC shape. Moreover, an open-class classification problem was solved using RBC scattering images with new shapes and landscape images.