Spectral characterization of rank filters based directional textures of digital images using Rajan transform

Abstract

Tissue characterization with the help of ultrasound images has remained an unsolvable problem to clinicians till date. Many techniques have been suggested to solve this issue. Yet a complete solution has not been arrived at so far. This paper gives a new technique which would indeed lead to the formulation of a robust method for characterizing tissues from ultrasound images. Any given image is processed using what we call as rank filters which would detect textures in four different directions. Various spatial features of these textures such as corners, curves, dots and lines are detected independently using the spectral domain pattern recognizing capabilities of Rajan Transform, which is a homomorphic transform developed on the lines of Hadamard Transform. The histogram analysis of these features would finally lead to spectral characterization of tissue textures. Clinicians would be able to resolve then the problem of tissue characterization. © 2012 ICST Institute for Computer Science, Social Informatics and Telecommunications Engineering.