Self-Similarity Driven Color Demosaicking

Abstract

Abstract-Demosaicking is the process by which from a matrix of colored pixels measuring only one color component per pixel, red, green or blue, one can infer a whole color information at each pixel. This inference requires a deep understanding of the interaction between colors, and the involvement of image local geometry. Although quite successful in making such inferences with very small relative error, state of the art demosaicking methods fail when the local geometry cannot be inferred from the neighboring pixels. In such a case, which occurs when thin structures or fine periodic patterns were present in the original, state of the art methods can create disturbing artifacts, known as zipper effect, blur, and color spots. The aim of this paper is to show that these artifacts can be avoided by involving the image self-similarity to infer missing colors. Detailed experiments show that a satisfactory solution can be found, even for the most critical cases. Extensive comparisons with state of the art algorithms will be performed on two different classic image databases.

Dictionary Learning for Noisy and Incomplete Hyperspectral Images

“…Hence, the model imposes a self-consistency in the use of dictionary elements, which is well matched to the properties of many natural images [46]. This is a key property of this sparseness construction, which is not accounted for in the Bayesian Lasso model in (4).…”

Color demosaicking by local directional interpolation and nonlocal adaptive thresholding

“…The color reproduction quality depends on the image contents and the employed CDM algorithms. 2 Various CDM algorithms [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18] have been proposed in the past decades. The classical second order Laplacian correction 3,4 (SOLC) algorithm is one of the benchmark CDM schemes due to its simplicity and efficiency.…”

“…The classical second order Laplacian correction 3,4 (SOLC) algorithm is one of the benchmark CDM schemes due to its simplicity and efficiency. The recently developed methods include the successive approximation-based CDM by Li, 9 the adaptive homogeneity CDM by Hirakawa et al, 10 the directional linear minimum mean square-error estimation (DLMMSE)-based CDM method by Zhang et al, 12 the directional filtering and a posteriori decision CDM by Menon et al, 13 the sparse representation-based method by Mairal et al, 14 and the nonlocal means-based self-similarity driven (SSD) method by Buades et al, 15 etc. A recent review of CDM methods can be found in Ref.…”

“…It was pointed out in Refs. 15,16, and 19 that images in the Kodak dataset have much higher spectral correlation, lower color saturation, and smaller chro-matic gradients than images in other datasets, e.g., the McMaster dataset used in this paper (refer to Sec. 3.1 for more information).…”

“…[21][22][23][24][25][26] The mathematical framework of NLM denoising was well established by Buades et al, 21 where the given pixel x is estimated as the weighted average of all pixels whose Gaussian neighborhoods look like the neighborhood of x. The recently developed SSD algorithm by Buades et al 15 is an NLM-based CDM scheme. In Sec.…”