Skip to Main content Skip to Navigation
Theses

Sur une approche de l'analyse en composantes indépendantes à la compression des images multi composantes

Abstract : In the first part of this work, we define several compression schemes based on JPEG2000 using two different transforms to reduce the spatial and spectral redundancies. The first one is the DWT (Discrete Wavelet Transform) that is well known with his properties to reduce in a good way the spatial redundancy between the wavelet coefficients. We develop the criterion to be minimized in order to obtain the optimal spectral transform of those compression schemes. This criterion equals the criterion used in independent component analysis (ICA) (when the mutual information is minimized) added with another term that could be seen as a measure of the transformation matrix to the orthogonality. Indeed, this additional term is always positive and becomes equal to zero if and only if the linear transform obtained is orthogonal. The results show that the new transforms perform better than the KLT (Karhunen Loève transform) that is usually used in compression to reduce the spectral redundancy. In the second part of our work, we define one new criterion in compression using a separation deconvolution modeling. This criterion is simply linked to the criteria that is minimized in ICA (when using mutual information like criterion to minimize) with a relation closed to that of the first part. We have evaluated the gradient and the Hessian of that criterion and choose to use a BFGS algorithm for the minimization of the criteria. Two algorithms were settled, the first one that minimizes the criteria in the general case and the second one, that minimizes the criteria with a relative simple constraint.
Complete list of metadatas

Cited literature [86 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00151987
Contributor : Isidore Paul Akam Bita <>
Submitted on : Tuesday, June 5, 2007 - 6:03:28 PM
Last modification on : Thursday, November 19, 2020 - 1:01:01 PM
Long-term archiving on: : Thursday, April 8, 2010 - 7:01:25 PM

Identifiers

  • HAL Id : tel-00151987, version 1

Collections

Citation

Isidore Paul Akam Bita. Sur une approche de l'analyse en composantes indépendantes à la compression des images multi composantes. domain_other. Université Joseph-Fourier - Grenoble I, 2007. Français. ⟨tel-00151987⟩

Share

Metrics

Record views

800

Files downloads

898