| Titre : | Segmentation des Images au Sens du Mouvement (Image Segmentation Basedon Motion) |
| Auteurs : | amina Ourchani, Auteur ; zine Eddine Baarir, Directeur de thèse |
| Type de document : | Monographie imprimée |
| Editeur : | Biskra [Algerie] : Université Mohamed Kheider, 2019 |
| Langues: | Anglais |
| Mots-clés: | Segmentation,moving object,background,foreground,color feature,texture feature,shapefeature,motion feature,NMES,MSFs,BMFS-LBP |
| Résumé : | Detection of moving objects from background in video sequences is hard, but essentialtask in a large number of applications in computer vision. Most of the existing methods hadgiven accepted results only in the case where both object and background are rigid, becauseof the serious occlusions and the complex computation, which presents limitations in case ofocclusions and shadows.In this thesis ,we developed three new approches to detect moving foregrounds from com-plex backgrounds . In the first approach called new method for the motion estimation andsegmentation of moving objects ”NMES”, we focus on the combination of motion, color andtexture features.Firstly, we have used the block-matching method to compute the motion vectorand we havetaken into our consideration the result of the frame difference technique, to improve the qualityof the optical flow. Moreover, we have used the k-means clustering algorithm owing to groupthe pixels, having similar motion, color and texture features.lastly, the results of grouping pixels are used as an input in Chan-Vese model, in order toattract the evolving contour of moving object contours.In the second approach called combination between motion and shape features ”MSFs”, weapplied a logical comparison between the results of the optical flow (motion feature) and thecolor space segmentation (shape feature) of each pixel.In the third approach called hybridization between motion and texture features ”BMFS-LBP”, we concentrated on a combination between local binary pattern (texture feature) andblock matching full Search algorithm (motion feature).Both of MSFs and BMFS-LBP approaches are suitable to discriminate moving objects fromboth static and dynamic background.Finally, to evaluate the performance of our proposed approaches, we experimentd themon challenging sequences. Ws have shown that our BMFS-LBP approach provided improvedsegmentation results compared with MSFs method and state of art methods. |
| Sommaire : |
Contents1 Introduction21.1Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .32 Chapter II. State of art of segmentation related to moving objects52.1Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .72.2Importance of moving objects segmentation . . . . . . . . . . . . . . . . . . .72.3Challenges and Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .92.4Traditional Motion Segmentation Techniques . . . . . . . . . . . . . . . . . .112.5Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .212.6Ground-truth ”gold standard” . . . . . . . . . . . . . . . . . . . . . . . . . . .282.7Performance Measure Evaluation Methodology . . . . . . . . . . . . . . . . .292.8conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .303 Chapter III. Review of different techniques related to feature extraction and seg-mentation323.1Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .333.2Feature extraction techniques . . . . . . . . . . . . . . . . . . . . . . . . . . .333.3Clustering algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .443.4Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .464 Chapter IV. Moving object segmentation based on feature extraction with staticbackground474.1Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .494.2Features extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .494.3Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .564.4Results and discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .574.5Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .725 ChapterVI. Moving object segmentation from complex videos based on featureextraction735.1Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .755.2Combination between motion and shape features (Combination Between Mo-tion and Shape Features (MSFs)) . . . . . . . . . . . . . . . . . . . . . . . . .765.3Hybridization between motion and texture features (Hybridization between Mo-tion and Texture Features (BMFS-LBP)) . . . . . . . . . . . . . . . . . . . . .805.4Experimental results and discussions . . . . . . . . . . . . . . . . . . . . . . .835.5CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .92vii
6 Chapter V: CONCLUSIONS AND FUTURE WORKS936.1Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .946.2Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9 |
Disponibilité (1)
| Cote | Support | Localisation | Statut | Emplacement | |
|---|---|---|---|---|---|
| TH/1026 | Thèse de doctorat | BIB.FAC.ST. | Empruntable |
Erreur sur le template
