PSNR & Eigen value based rust defect recognition & evaluation of steel coating conditions

Abstract: PSNR is one of the most often and universally used method for measuring quality of image. In this paper we propose a methodology for assessment of coating condition of bridge images. The defect recognition algorithm includes conversion of captured images into grey level; these grey level images are grouped into defective & non defective group. Further that is processed to plot correspondence map. The correspondence map is measure of matching image. Straight line with 450 in correspondence map indicates no defect in scene image. In contrast if correspondence map produces nonlinear image it indicates defect (rust) in scene image. The nonlinear shape of grey level distribution in correspondence map can be analyzed by calculating Eigen values. Two similar images will produce smaller Eigen value (approximately zero), whereas it will be distinctly large for dissimilar images. The PSNR determines proportion of rust in scene image with relation to reference image.

Keywords: Rust detection, covariance matrix, Eigen values, correspondence map, PSNR, MSE.