| 摘要 |
An adaptive filter according to one embodiment includes an NxN two-dimensional convolution, where N is an odd integer greater than one. Pixels of an image are filtered by moving the NxN kernel across the image. At each pixel location, the kernel coefficients are determined by the image pixels within the NxN region, such that the kernel coefficients may change from one pixel location to the next. Such a filter may be implemented to use a lookup table to approximate a floating-point operation. |
