Synthetic Results: Mountain Scene
PSNR of each frame in the input sequence and in the results of the 3 methods in comparison. Our method performs best in terms of preserving texture details.
Frame 23
Please move your mouse cursor over the small images to toggle between the results.
| Constant exposure (1/60) |
Noisy input PSNR = 25.96 |
CBM3D PSNR = 32.39 |
Liu and Freeman PSNR = 32.67 |
Our algorithm PSNR = 39.27 |
Ground truth |
|---|---|---|---|---|---|
|
|
|
|
|
|
Constant exposure time |
|||||
Liu and Freeman's method over-smooths the grass and tree branches significantly. CBM3D retains more texture and detail. Our result is most comparable to the ground truth. Difference images with respect to the ground truth are shown below for the full frame.
| Ground truth | CBM3D | Liu and Freeman | Our algorithm |
|---|---|---|---|
|
|
|
|
Ground truth image |
|||
Frame 31
| Constant exposure (1/60) |
Noisy input PSNR = 23.28 |
CBM3D PSNR = 32.44 |
Liu and Freeman PSNR = 32.63 |
Our algorithm PSNR = 36.47 |
Ground truth |
|---|---|---|---|---|---|
|
|
|
|
|
|
Constant exposure time |
Our algorithm is the best of the three at preserving texture; the other two denoising algorithms over-smooth the details in the trees. Difference images are shown below for the full frame.
| Ground truth | CBM3D | Liu and Freeman | Our algorithm |
|---|---|---|---|
|
|
|
|
Ground truth image |
|||
Video
| Top left: A noisy input video captured using motion-based exposure control. | Top right: CBM3D denoising result. |
| Bottom left: Liu and Freeman denoising result. | Bottom right: Our denoising result. |
The differences between these results are hard to notice in the presence of motion and strong video compression. For applications that require extracting individual frames from the video, the advantages of our algorithm are more pronounced, as shown above using frames 23 and 36.