PROJECT TITLE :

Quadtree Structured Image Approximation for Denoising and Interpolation - 2014

ABSTRACT:

The success of the many image restoration algorithms is often thanks to their ability to sparsely describe the first signal. Shukla proposed a compression algorithm, based mostly on a sparse quadtree decomposition model, that may optimally represent piecewise polynomial images. During this paper, we tend to adapt this model to the image restoration by changing the speed-distortion penalty to a description-length penalty. In addition, one of the major drawbacks of this sort of approximation is that the computational complexity required to find a appropriate subspace for every node of the quadtree. We have a tendency to address this issue by searching for a suitable subspace abundant a lot of efficiently using the mathematics of updating matrix factorisations. Algorithms are developed to tackle denoising and interpolation. Simulation results indicate that we tend to beat state-of-the-art results when the initial signal is in the model (e.g., depth images) and are competitive for natural images when the degradation is high.