| 摘要 |
<p>Input image signals, to which noise is added on a transmission line or the like, are fed, and output image signals, from which the noise is eliminated by a classification adaptive processing, are outputted. In the classification adaptive processing, prediction coefficients are determined in advance for each class by learning to determine a class reflecting the noise components in the input image signals, and a predicted pixel value is generated by a linear combination between the prediction coefficients of the class and the values of a plurality of pixels contained in a plurality of frames including the pixel of concern in the input image signals. The predicted pixel value is free from influence of noise. The motion of the pixel of concern is detected, and motion the pixels used for determining the class and the pixels used for the prediction operations are subjected motion correct thereby reflecting the noise components accurately.</p> |
