Sony Corporation

Tomonori Kawase

Kunio Kawaguchi

Tetsujiro Kondo

Yasuaki Takahashi

Takafumi Morifuji

Takeharu Nishikata

An image processing method includes determining whether an image is corrupted, extracting class taps used for classification of a pixel to be generated, performing classification of the pixel to be generated by using the class taps, extracting prediction taps used for prediction of the pixel to be generated, and generating an image by predictively calculating a pixel value of the pixel to be generated by using the prediction taps and prediction coefficients for the class of the pixel among a first prediction coefficient set if the subject pixel is determined as a pixel representing a corrupted image, and by predictively calculating a pixel value of the pixel to be generated by using the prediction taps and prediction coefficients for the class of the pixel to be generated among a second prediction coefficient set if the subject pixel is determined as a pixel representing an uncorrupted image.

Image processing apparatus and method, learning apparatus and method, and program

A compensation allocation unit performs allocation compensation of a motion vector to a pixel to which a motion vector is not allocated by a victor allocation unit. A C/UC area determination unit compares magnitudes of a present DFD, a past DFD, and a future DFD based on a background vector and a DFD based on a compensation allocation vector by an evaluation value calculation unit to determine an area of a target pixel. Regarding the target pixel determined as the pixel in a covered area or an uncovered area, an interpolation method decision unit decides a computation method for a pixel value through which one of a double-sided interpolation and a one-sided interpolation. An interpolation frame generation unit computes the target pixel of the pixel value of the interpolation frame through the method decided by the interpolation method decision unit.

Growing community of inventors

Tokyo, Japan

Tomonori Kawase