We investigate the implications of a unified spatio-chromatic basis for image compression and reconstruction. Different adaptive and general methods (principal component analysis, PCA, independent component analysis, ICA, and discrete cosine transform, DCT) are applied to generate bases. While typically such bases with spatial extent are investigated in terms of their correspondence to human visual perception, we are interested in their applicability to multimedia encoding. The performance of the extracted spatio-chromatic spatial patch bases is evaluated in terms of quality of reconstruction with respect to their potential for data compression. Since ICA is not as widely used as it should be, compared to the other decorrelation methods applied here in a new domain, we also provide a review of ICA. The results discussed here are intended to provide another path towards perceptually based encoding of visual data. This leads to a deeper understanding of the role played by chromatic features in data reduction.

/pdf/spatio-chromatic-decorrelation-for-color-image-compression-3uvrja20e5.pdf

Spatio-chromatic decorrelation for color image compression

https://hal-supelec.archives-ouvertes.fr/hal-00278351/document

ICA based algorithms for computing optimal 1-D linear block transforms in variable high-rate source coding

In this paper we propose some compression schemes for multicomponent satellite images. These compression schemes use a classical bi-dimensional discrete wavelet transform (DWT) for spatial redundancy reduction, associated with linear transforms that reduce the spectral redundancy in an optimal way, for uniform scalar quantizers. These transforms are returned by Independent Component Analysis (ICA) algorithms which have been modified in order to maximize the compression gain under the assumption of high rate quantization and entropy coding. One algorithm, called ICA_opt, returns an optimal asymptotical linear transform and the other, called ICA_orth, returns an optimal asymptotical orthogonal transform. We compare the performance in high and medium rate coding of the Karhunen Loeve Transform (KLT) with the transforms returned by the modified ICA algorithms. These last transforms perform better than the KLT in term of compression gain in all cases, and in some cases the gain becomes significant.

Compression of multicomponent satellite images using independent components analysis

Our research focuses on reducing complexity of hyperspectral image codecs based on transform and/or subband coding, so they can be on-board a satellite. It is well-known that the Karhunen-Loeve Transform (KLT) can be sub-optimal in transform coding for non Gaussian data. However, it is generally recommended as the best calculable linear coding transform in practice. Now, the concept and the computation of optimal coding transforms (OCT), under low restrictive hypotheses at high bit-rates, were carried out and adapted to a compression scheme compatible with both the JPEG2000 Part2 standard and the CCSDS recommendations for on-board satellite image compression, leading to the concept and computation of Optimal Spectral Transforms (OST). These linear transforms are optimal for reducing spectral redundancies of multi- or hyper-spectral images, when the spatial redundancies are reduced with a fixed 2-D Discrete Wavelet Transform (DWT). The problem of OST is their heavy computational cost. In this paper we present the performances in coding of a quasi optimal spectral transform, called exogenous OrthOST, obtained by learning an orthogonal OST on a sample of superspectral images from the spectrometer MERIS. The performances are presented in terms of bit-rate versus distortion for four various distortions and compared to the ones of the KLT. We observe good performances of the exogenous OrthOST, as it was the case on Hyperion hyper-spectral images in previous works.

Lossy compression of MERIS superspectral images with exogenous quasi optimal coding transforms

Les travaux developpes dans cette these portent sur les decompositions multiresolution dans un cadre de lifting scheme, appliquees a la compression d'images. Pour commencer, une structure de decomposition consistant en un “lifting scheme generalise” est proposee. Ce schema permet d'exploiter toute l'information disponible au decodage dans l'etape de prediction. Cela est rendu possible par l'ajout d'un filtre de prediction supplementaire par rapport a la structure classique de lifting scheme. Le schema propose est ensuite associe a deux methodes d'adaptation. La premiere, appelee GAE, suppose une stationnarite globale de l'image, tandis que la seconde, LAE ne suppose qu'une stationnarite locale de l'image. Dans ce dernier cas, les filtres predicteurs sont adaptatifs. Trois applications de ces methodes en compression d'images sont ensuite proposees. Dans un premier temps, une comparaison des performances en compression sans perte sur des images de textures, synthetiques, gaussiennes, a stationnarites locale et globale (verifiant les hypotheses plus haut), est realisee. Sur ces signaux, les mesures d'entropie d'ordre 1 ont montre que les methodes proposees offrent en moyenne un gain en codage de 0,5 bpp (GAE) et 0,8 bpp (LAE) par rapport a la decomposition en ondelette (9,7), de 0,8 bpp (GAE) et 1,11 bpp (LAE) par rapport a la (5,3) et de 0,41 bpp (GAE) et 0,65 bpp (LAE) par rapport a la methode de Gerek et Cetin. La deuxieme application concerne le codage sans perte d'images reelles de natures variees. Les gains par rapport a l'etat de l'art se sont reveles plus faibles que ceux obtenus pour les images synthetiques. Enfin, la derniere application traite les cas du codage progressif et du codage avec perte. Pour la compression avec pertes, nous avons modifie la methode LAE pour palier aux problemes de divergence dus a l'impossibilite au niveau du decodeur de reconstruire les filtres predicteurs a partir d'echantillons quantifies. Elle se revele plus efficace que lorsque l'on utilise les filtres usuels de longueur fixe (9,7) et (5,3).

Synthèse de bancs de filtres adaptés, application à la compression des images.

In this paper we present two new algorithms that compute the linear optimal transform in high-rate transform coding, for non Gaussian data. One algorithm computes the optimal orthogonal transform, and the other the optimal linear transform. Comparison of the performances in high-rate transform coding between the classical Karhunen-Loeve Transform (KLT) and the transforms returned by the new algorithms are given. On synthetic data, the transforms given by the new algorithms perform significantly better that the KLT, however on real data all the transforms, included KLT, give roughly the same coding gain.

I.P. Akam-Bita

Papers

Modified ICA algorithms for finding optimal transforms in transform coding