Abstract:
The article investigates the problem of restoring a raster image on defective areas of CCD or CMOS matrices. The task of image restoration is divided into subtasks: restoration of contours and texture restoration. These problems are solved by a special image recovery machine, which uses image classification methods adapted to the task, a neural network, and image recovery methods. An original method of image restoration in the video series is proposed. The analysis of the computational complexity of the methods is fulfilled.
Keywords:texture generation, discrete function approximation, image statistics, Haralik and Laws parameters, Gabor filter, neural networks.