Image processing device, learning device, and coefficient generating device and method

Abstract

A target-pixel-setting section () sets a target pixel in a target image to be predicted. A motion-direction-detecting section () detects a motion direction corresponding to the target pixel. A pixel-value-extracting section () extracts from peripheral images corresponding to the target image, in order to extract a main term that mainly contains component of the target pixel in a moving object that encounters a motion blur in the peripheral images, at least pixel values of pixels in the peripheral images whose space position roughly agree with space position of the target pixel. A processing-coefficient-setting section () sets a specific motion-blur-removing-processing coefficient. A pixel-value-generating section () newly generates pixel values for processing from the pixel values extracted by the pixel-value-extracting section () corresponding to the motion direction and generates pixel values corresponding to the target pixel based on the pixel values for processing and the specific motion-blur-removing-processing coefficients. It is possible to perform a robust motion-blur-removing processing on any shifts of motion vector.