An RGB-Z system is a vision system-on-chip that captures both color and depth information. Recently, the idea of combining color and depth acquisition at the pixel level and in the same array has emerged. Such a sensor would allow a more compact integration while reducing the calibration problems compared to a system composed of two sensors. In this thesis, we focus on the problems of reconstructing the missing color information due to the heterogeneity of the pixel matrix. In a first step, we proposed different RGB-Z matrices. In a second step, several algorithms for the reconstruction of the missing information inspired by the state of the art have been adapted and implemented. Moreover an original adaptive algorithm to improve the quality of the reconstructed images is presented. Finally, a first hardware implementation of the algorithm is proposed.