My research works are globally turned toward computational photography and computer vision. Especially, I am interested in both image editing and synthesis. During my PhD I have worked on image and video inpainting applied to stereoscopic data with depth maps. This research was conducted within a French national project whose aim was to conceive solutions for virtual view synthesis in 3D movies theaters. First, we have enhanced the state-of-the-art greedy inpainting algorithm of Criminisi et al. by proposing a smarter way to search patches in the image together with a tensor-based data term that improves the way to order the image reconstruction. Then, we have proposed an artifact reduction method for image and video, that improves the results of any greedy inpainting algorithm by removing the block-effect they contain. To do so we proposed an isotropic method, and then improved the latter by creating an anisotropic tensor model to guide the blending process.