Curved Surface Feature Extraction and Texture Analysis Based on Curvelet Transform for Red Blood Cell Image

Abstract

The surface feature of erythrocyte image plays an important role in clinic blood-related diagnosis. In this paper, a methodology was proposed firstly to extract the curved surface feature which started with image preprocessing. Afterwards the mean curvature and Gaussian curvature were computed through quadratic surface fitting with discrete data points, which exhibited the surface type distribution of different pixel regions. According to the irregular curved shape distribution properties of the RBC image, we put forward that it will make sense if we take the curved surface feature into account with the curvelet transform, which should provide meaningful information to extract the texture feature to classify the red blood cell categories.