Missing Data Correction in still images using Multi-Resolution Analysis

This paper proposes an improvement to the texture generation-based image in-painting algorithm, using multi-resolution analysis. Instead of optimizing the size of the neighborhood window of the synthesized pixel and the size of the test image, the approach described in this paper applies the same (minimal) window and test image size to different sub-bands of the discrete wavelet transformed (DWT) image. This way the execution complexity is kept lowest possible, while still obtaining the same visual quality as by applying the optimal parameters in the spatial domain.