In this study, a simple technique for defect identification based on Orthogonal Polynomials (OP) model is presented. Initially, the input image under analysis is applied with OP model and gradient estimation scheme is employed to locate the edges present. The resulting binary image is again applied with OP model, and a simple computation scheme that finds the ratio between selected transform coefficients is proposed to identity the defects present in the image. Experiments have been conducted with different images consisting of both homogeneous and non homogeneous regions. The proposed technique is found to perform well, for unshaped defects, and is found to outperform the existing schemes.
展开▼
