The classical definition of vector order filters consists in selecting the pixel that minimizes the cumulated distance to the other pixels of the filtering window. The "most representative" pixel of the filtering window is then... more
The classical definition of vector order filters consists in selecting the pixel that minimizes the cumulated distance to the other pixels of the filtering window. The "most representative" pixel of the filtering window is then selected and the noise is thus reduced. But, when the filtering window reaches a transition, the result is biased and the smoothing is not optimal any more. The proposed technique consists in progressively decimating the filtering window by suppressing the pixels that maximize the cumulated distance. The process is then iterated (computation of the cumulated distances among the remaining pixels and suppression of the less typical pixels) until one single pixel remains. It is then selected for the filter output. In practical cases, the number of iterations is limited. Applied on colour images, the filter results in an enhancement of the edges and in an automatic registration of the different components. The counterpart is a slight degradation of the smoothing performance in homogeneous regions. The paper details the proposed method and compares it with other enhancement filters.
High resolution images provided by synthetic aperture sonar (SAS) sensors are of great interest, especially for the detection, location and classification of mines lying on the sea bed. But these data obtained by an active imagery system... more
High resolution images provided by synthetic aperture sonar (SAS) sensors are of great interest, especially for the detection, location and classification of mines lying on the sea bed. But these data obtained by an active imagery system are highly corrupted by a noise called the speckle. To reduce this noise and suppress the spurious reflections it generates on the images,
Research Interests:
In this paper, a total ordering scheme for multivariate data , based on the bit mixing paradigm, is presented. This ranking scheme is then used to extend morphologi- cal filters to the vectorial case. Results concerning the al- ternating... more
In this paper, a total ordering scheme for multivariate data , based on the bit mixing paradigm, is presented. This ranking scheme is then used to extend morphologi- cal filters to the vectorial case. Results concerning the al- ternating sequential filtering by reconstruction of colour images are presented.
Research Interests:
In this letter, we propose a method aiming at reducing the noise in hyperspectral images based on the nonlinear generalization of principal component analysis (NLPCA). NLPCA is performed by an autoassociative neural network (AANN) that... more
In this letter, we propose a method aiming at reducing the noise in hyperspectral images based on the nonlinear generalization of principal component analysis (NLPCA). NLPCA is performed by an autoassociative neural network (AANN) that has the hyperspectral image as input and is trained to reconstruct the same image at the output. Due to its topology, characterized by a bottleneck layer, the nonlinear AANN forces the hyperspectral image to be projected in a lower dimensionality feature space by removing noise and both linear and nonlinear correlations between spectral bands. This process permits to obtain enhancements in terms of the quality of the reconstructed hyperspectral image. The results conducted on different hyperspectral images are qualitatively and quantitatively discussed and demonstrate the potentialities of the proposed method, as compared with similar approaches such as PCA and kernel PCA.
Research Interests:
... {fredericmaussang ; jocelyn.chanussot}@lis.inpg.fr ... C. Collet, P. Thourel, M. Mignotte, P. Perez, and P. Bouthemy, Segmentation markovienne hierarchique multimodkle d'images sonar haute resolution., Traitement du Signal,... more
... {fredericmaussang ; jocelyn.chanussot}@lis.inpg.fr ... C. Collet, P. Thourel, M. Mignotte, P. Perez, and P. Bouthemy, Segmentation markovienne hierarchique multimodkle d'images sonar haute resolution., Traitement du Signal, vol. 15, n3, pp. 231 - 250, 1998. ...
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Synthetic Aperture Sonar (SAS) imagery is largely used in de- tection, location and classification of underwater mines laying or buried in the sea bed. This paper proposes a detection method us- ing Higher Order Statistics (HOS) on SAS... more
Synthetic Aperture Sonar (SAS) imagery is largely used in de- tection, location and classification of underwater mines laying or buried in the sea bed. This paper proposes a detection method us- ing Higher Order Statistics (HOS) on SAS images. The proposed method can be divided into two steps. Firstly, the HOS (Skewness and Kurtosis) are locally estimated using a square sliding compu- tation window. In a second step, the results are focused by a cor- relation process. This enables the precise location of the objects. This method is tested on real SASdata containing both underwater mines laying on the sea bed and buried objects.