| 摘要 |
The transposed filter of the image degrading filter is calculated, and a residual gradient image is calculated by the convolution operation of the transposed filter and a residual image obtained from the image degrading filter, the estimated restored image and the degraded image. Then, a modified image of the estimated image is calculated from the residual gradient image, and the estimated image is modified and updated based on the modified image. Consequently, even when the degradation function is one that can be expressed as a complicated helical filter, the memory consumption is small and the calculation time is short compared to a method depending on a large matrix and an inverse matrix.
|
