Picture resolution enhancement
Ernst Boyd, Kiyoshi Akama, Arthur W. Brody · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1997
When making an animation by moving inside a 3D scene, areas close to the camera will be larger on the screen and the limited resolution of the data will become visible. To make more natural appearing scene it is necessary to enhance the resolution of these areas. An initial version is made from the lower resolution data by interpolation and then individual pixels are adjusted to make it better match a desired texture. The texture is derived from a number of closeup photographs by a classification algorithm which compares the local low resolution data with low resolution versions of the closeups. Rather than choosing a single closeup as a target texture, the classification algorithm generates a similarity score between thee region to be enhanced and each of the closeup samples. This score is then sued as a weighting function to determine how closeups may be used directly by comparing them to the enhanced region as it is being corrected, or for increased efficiency, they may be measured for some parameters and then the enhanced region is made to match these parameters.