An automatic transformation from bimodal to pseudo-binary images

Abstract

In this paper a procedure is proposed to transform bimodal images into enhanced ones with unified properties. These (pseudo-binary) images keep valuable grey level information and can be easily segmented by a global thresholding technique, using always the same threshold in the middle of the grey scale. This transformation is automatically self-adjusted from the statistical characterization of the histogram modes. It can be useful for the automatic segmentation of bimodal images found in controlled illumination environments, dealing with possible uncontrolled light variations. In addition, the grey level values in the pseudo-binary image can be considered as occupation percentage of the pixels, so subpixel reasonings can be easily inferred from the data in this new image.