Clustering-Based Denoising With Locally Learned Dictionaries

Abstract

Abstract-In this paper, we propose K-LLD: a patch-based, locally adaptive denoising method based on clustering the given noisy image into regions of similar geometric structure. In order to effectively perform such clustering, we employ as features the local weight functions derived from our earlier work on steering kernel regression [1]. These weights are exceedingly informative and robust in conveying reliable local structural information about the image even in the presence of significant amounts of noise. Next, we model each region (or cluster)-which may not be spatially contiguous-by "learning" a best basis describing the patches within that cluster using principal components analysis. This learned basis (or "dictionary") is then employed to optimally estimate the underlying pixel values using a kernel regression framework. An iterated version of the proposed algorithm is also presented which leads to further performance enhancements. We also introduce a novel mechanism for optimally choosing the local patch size for each cluster using Stein's unbiased risk estimator (SURE). We illustrate the overall algorithm's capabilities with several examples. These indicate that the proposed method appears to be competitive with some of the most recently published state of the art denoising methods.Index Terms-Clustering, dictionary learning, image denoising, kernel regression, principal component analysis, Stein's unbiased risk estimator (SURE).

Nonlocal Hierarchical Dictionary Learning using Wavelets and Gradient Histogram Preservation for Image Denoising: A Review

“…Therefore, for wavelet patches from the same scale, propose to train different sub-dictionaries in different underlying clusters. Then simply to obtain a light representation in the form that similar wavelet patches use the same dictionary elements but with different sparse codes [2], [3], [4].This would improve the coding efficiency and reduce artifacts. In NHDLWD, Use clustering to obtain training samples for each sub-dictionary.…”

A New Approach to Image Denoising by Patch-Based Algorithm

“…It is proved that regional M-smoothing uses the initial image in the averaging procedure and determine the minimum of a local criterion considering that iterated bilateral filtering uses the evolving image and must stop subsequently a particular number of iterations in order to avoid a flat image [15]- [20].…”

“…This transformation method is a moderately change-point detection procedure introduced by Lepskii [33]. The recommended procedure shares some common points with the contemporary non-local means algorithm [17], other patch -based methods [15], [18], [29] and the DUDE algorithm. In [16], the authors design a two-pass approach and substitute the most frequent patch /symbol seen in a local window to the current perverted patch /symbol.…”

E-PLE: an Algorithm for Image Inpainting

“…The orientation based K-LLD for image denoising [5] is another example. In a paper by Yu et al [15] a similar algorithm, called PLE, was designed but intended to solve generic image related inverse problems.…”

“…In this contribution, motivated in part by the works of Chatterjee et al [5], Yu et al [15], and Zoran et al [16], we present E-PLE, or Enhanced PLE. Using a specialized Gaussian mixture initialized with real-world images, we adapt expectation maximization (EM) algorithm [7] to this particular setting and show its improved performance at inpainting.…”