|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| Test images |
From here, we can see that when the noise is small, the estimation is not accurate enough. This is due to the estimation error which corresponds to the weak edges of the image. This error is actually equal the estimation on the noise-free image. It is usually very small. When increasing the noise level, this error will become less and less significant. The third column refers to the standard deviation of the estimation. The values are pretty small, which means this technique is quite robust and has low dependency on the image under estimation.
The following are the estimation results on the uniform noise. This method also provides good performance.
