Color Converting of Endoscopic Images Using Decomposition Theory and Principal Component Analysis - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2019

Color Converting of Endoscopic Images Using Decomposition Theory and Principal Component Analysis

(1, 2) , (1) , (1) , (1)
1
2

Résumé

Endoscopic color imaging technology has been a great improvement to assist clinicians in making better decisions since the initial introduction. In this study, a novel combined method, including quadratic objective functions for the dichromatic model by Krebs et al. and Wyszecki`s spectral decomposition theory and the well-known principal component analysis technique is employed. New algorithm method working for color space converting of a conventional endoscopic color image, as a target image, with a Narrow Band Image (NBI), as a source image. The images of the target and the source are captured under known illuminant/sensor/filters combinations, and matrix Q of the decomposition theory is computed for such combinations. The intrinsic images which are extracted from the Krebs technique are multiplied by the matrix Q to obtain their corresponding fundamental stimuli. Subsequently, the principal component analysis technique was applied to the obtained fundamental stimuli in order to prepare the eigenvectors of the target and the source. Finally, the first three eigenvectors of each matrix were then considered as the converting mapping matrix. The results precisely seem that the color gamut of the converted target image gets closer to the NBI image color gamut.

Dates et versions

hal-02487573 , version 1 (21-02-2020)

Identifiants

Citer

Keivan Ansari, Alexandre Krebs, Yannick Benezeth, Franck Marzani. Color Converting of Endoscopic Images Using Decomposition Theory and Principal Component Analysis. 9th International Conference on Computer Science, Engineering and Applications, Jul 2019, Toronto, Canada. pp.151-159, ⟨10.5121/csit.2019.91812⟩. ⟨hal-02487573⟩
26 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook Twitter LinkedIn More