The mathematical/statistical ideas used to address two or three applications of computational vision will be discussed. In the restoration of motion pictures, film is scanned, restored or repaired by computer, and either transferred to a video format or returned to film. Software has been developed for automatic removal of scratches and dirt and for correction of other defects. There are many problems of mathematical or statistical nature, including motion estimation, two and three-dimensional signal interpolation, and synthesis of textures. In a second application, to industrial automation, manufactured items are tracked by automatically reading optically-imaged identification characters. The characters occurring in the industrial environments are typically severely degraded by effects of manufacturing processes. A PC-based algorithm uses ideas nonparametric statistics to accommodate the drastic variations resulting from marking and imaging conditions and uses a hierarchical decision strategy to infer character IDs from degraded image data.