Depth Map Super-Resolution for Cost-Effective RGB-D Camera

Recently, cost-effective RGB-D cameras measuring distance to target objects have been used in many fields. By using a depth camera equipped with the RGB-D camera, we can easily recognize depth and three-dimensional objects having difficulty in distinction only with RGB images. However, the depth cameras are low-resolution compared with the RGB cameras, and include measuring errors. Therefore, in this paper, we present a method to reduce the measurement errors of the depth camera and enhance the resolution of a depth map by using a high-resolution RGB camera image. Especially around object's contours, super-resolution depth maps are obtained by an interpolation process from the pixels that is selected by the color similarity and selection pixels stored reliable depth value. Then, we confirm that the depth maps are the same resolution as the RGB image, and are more similar to the objects shape indicates by the color image than only expanded depth map. Further, we confirm that result is improved by applying this approach to the application combined with high-resolution RGB image.