Information Technology Reference
In-Depth Information
(a)
(b)
(c)
(d)
(e)
Fig. 9.20. Noise removal and contrast enhancement: (a) hidden feature array in Layer 0; (e)
output feature array; contributions to the output activity (b) via input projections; (c) via
lateral projections (center-surround interaction); (d) via backward projections (indicating an
estimate of the background intensity and the line positions in higher layers).
25
TRN
TST
TRN
TST
1
20
15
0.1
10
0.01
5
0
0.001
5
10
15
20
5
10
15
20
(a)
(b)
iteration
iteration
Fig. 9.21. Noise removal and contrast enhancement - performance over time: (a) mean
squared error; (b) mean square delta output.
would be possible by considering isolated frames only. For instance, Elad and
Feuer [62] proposed a least squares adaptive filtering approach to increase the reso-
lution of continuous video.
Another example is the work of Kokaram and Godsill [127]. They proposed a
Bayesian approach to the reconstruction of archived video material. Using Markov
chain Monte Carlo methods, they simultaneously detect artifacts, interpolate miss-
ing data, and reduce noise.
9.5.1 Image Degradation
In the next experiment, the capability of the same Neural Abstraction Pyramid net-
work, used for the previous two tasks, to reconstruct digits from a sequence of de-
graded images is explored. The only difference is the image degradation procedure.
Search WWH ::




Custom Search