An Efficient Defect Estimation and Inpainting Based on Sparse Representation

Authors

  • M.Ganesh, M. Ravichandran and Sridhar Gummalla

Abstract

In this paper, a simple defect identification followed by efficient inpainting that compensates the missing details in defect images, with transform coefficients is presented. This proposed scheme initially decomposes the given defect image, coefficient-wise based on geometrical and textural primitives present in the difference between original and its Gaussian smoothed images. Then simple defect estimation is carried out based on  (i) strengthening the edge coefficients, and  (ii) location of transition between edge and texture primitives. The basic concept behind these procedures is the contribution of orthogonal polynomials model (OPM) coefficients as a sparse representation, towards low level primitives edge and texture. A simple structure inpainting is then employed with edge magnitude and orientation, for the defected edge coefficients. With homogeneity among orthogonal polynomials texture coefficients, a texture inpainting is then proposed with statistical analysis. The proposed inpainting scheme is evaluated with standard performance measures and compared with recent inpainting methods.

Published

2020-11-01

Issue

Section

Articles