Polarimetric Image Discrimination With Depolarization Mueller Matrix

Abstract

Polarimetric imaging techniques exploit the polarization characteristics of objects and, thus, can achieve a better discriminative performance. In this paper, an adaptive discrimination method based on the classification of depolarization Mueller matrix is proposed. According to the Mueller–Jones theory, the general formula of Mueller matrix for depolarization optical system is analyzed. In addition, a two-channel imaging platform is constructed to obtain a polarization-difference image with certain states of polarization. Under this methodology, every pixel of the image can be expanded as a combination of independent entries of the Mueller matrix, with polarization states as the representative coefficients. Thus, the optimal polarization states can be obtained via support vector machine (SVM), and a high contrast image is achieved. Finally, experiments on two groups of different materials are conducted to demonstrate the applicability and performance of the proposed method. The related criteria (e.g., Fisher ratio) are introduced to quantitatively evaluate the results. Experimental results indicate that the proposed method shows advantages for image discriminations.