Digital Signal Processing Reference
In-Depth Information
a) movie scene b) human scene
c) outdoor scene d) nature scene
Fig. 4.
Comparative results for CMMB Videos
only uses light model to eliminate shadow area in an image which can not get more
reference to some relevant information of the original image. Therefore, the result of
Retinex method can not give satisfied vision effect of an image with blindness
shadow.
5
Conclusions
In this paper, we present a solution for improving CMMB video's quality by remov-
ing shadow effects. The contributions of this paper can be summarized as follows:
(1) Geometry method is used to find similar features in high-dimensional space.
The method is blindness with no prior knowledge. The proposed method is robust.
(2) Combined with Gabor filters, we can use multi-scale analysis which can im-
prove speed. Unlike the previous methods, we do not make much assumption on sha-
dowsĂ properties.
Experimental results indicate our method is superior in terms of accuracy, robust-
ness, and stability in shadow elimination. The novelty of this solution is the direct
embedding of finding reflecting parameters of image using points in high dimensional
space. More studies in the aspect of basic theory in high-dimensional space inspired
by biology should be invested in future.
