Search

Additional depth information from RGBD images is one of characteristics different from conventional 2D images. In this paper, we propose an effective saliency model to detect salient regions in RGBD images. Color contrast and depth contrast are first enhanced with the weighting of depth-based object probability. Then the region merging based saliency refinement is exploited to obtain the color saliency map and depth saliency map, respectively. Finally, a location prior of salient objects is integrated with color saliency and depth saliency to obtain the regional saliency map. Both subjective and objective evaluations on a public RGBD image dataset demonstrate that the proposed saliency model outperforms the state-of-the-art saliency models.

A novel strategy of demosaicking and zooming for Color Filter Array via residual image reconstruction is presented. The proposed scheme depends on certain demosaicking and zooming techniques. Given demosaicking and zooming methods, a residual image between 'genuine' image and the initial demosaicked and enlarged image is reconstructed using a dictionary and sparse coding. The 'genuine' image has richer edges and details than that of the initial result image. The dictionary is obtained by learning a training image set and sparse representation. Once the residual image is reconstructed, scaled version of it is added back to the initial result image to get final result. As R-G difference and B-G difference are smoother than individual channels, we search G channel, R-G difference and B-G difference instead of individual three channels. Considering implementation efficiency, only 'genuine' G channel is reconstructed using the proposed method, the demosaicked R-G difference and B-G difference are just zoomed with simple bicubic interpolation. The experiment results have demonstrated the state-of-the-art results in both visual perception and PSNR.