PROJECT TITLE :

Progressive Image Denoising Through Hybrid Graph Laplacian Regularization A Unified Framework - 2014

ABSTRACT:

Recovering images from corrupted observations is critical for many real-world applications. In this paper, we tend to propose a unified framework to perform progressive image recovery based on hybrid graph Laplacian regularized regression. We have a tendency to initial construct a multiscale representation of the target image by Laplacian pyramid, then progressively recover the degraded image in the size house from coarse to fine thus that the sharp edges and texture can be eventually recovered. On one hand, inside every scale, a graph Laplacian regularization model represented by implicit kernel is learned, that simultaneously minimizes the least square error on the measured samples and preserves the geometrical structure of the image information area. In this procedure, the intrinsic manifold structure is explicitly considered using each measured and unmeasured samples, and therefore the nonlocal self-similarity property is utilised as a fruitful resource for abstracting a priori knowledge of the photographs. On the opposite hand, between two successive scales, the proposed model is extended to a projected high-dimensional feature house through specific kernel mapping to explain the interscale correlation, in that the local structure regularity is learned and propagated from coarser to finer scales. In this way, the proposed algorithm gradually recovers a lot of and a lot of image details and edges, which might not been recovered in previous scale. We have a tendency to check our algorithm on one typical image recovery task: impulse noise removal. Experimental results on benchmark check images demonstrate that the proposed methodology achieves higher performance than state-of-the-art algorithms.