Image Processing Reference
In-Depth Information
Figure 7.22
An example of removal of sensor noise caused by low lighting conditions. (a) shows
the image with sensor noise; (b) displays the filtered image.
It was found that for video sequences as opposed to still images it was neces-
sary to employ spatio-temporal filters to avoid motion artifacts.
7.6.3 Image deblurring
A further problem in film footage is that of blurring resulting from motion and from
autofocus cameras during panning when the distance to the subject has not stabi-
lized. It is very easy to create training data for blur problems simply by low-pass fil-
tering footage with sharp detail and using this as the noisy data with the original
footage as the ideal data.
Figure 7.23 shows an example of image deblurring with an aperture filter. The
aperture was trained on the deliberately blurred image of a lab. Figure 7.23(a)
shows a blurred image and Fig. 7.23(b) shows the same image after application of
the deblurring aperture. It can be seen that the clock is distinctly sharper.
