Bayesian Image and Video Enhancement Using a Non-Gaussian Prior - Nonlinear Signal and Image Processing: Theory, Methods and Applications - page 324

(a) (b)

(c) (d)

FIGURE 9.15
Superresolution video enhancement of the IKONOS satellite picture with an effective GSD of
4 m, using Bayesian estimation with the Huber-Markov random field prior model, T H = 1 . 5

λ ( i ) =

and 1000. (a) Microview of the original 1-m image data. (b) 4-m image data. (c) Bayesian

enhancement of the 4-m pixels using 1 frame. (d) Bayesian enhancement of the 4-m pixels using
3 frames. (e) Bayesian enhancement of the 4-m pixels using 6 frames. (f) Bayesian enhancement
of the 4-m pixels using all 16 available frames. Note that these nonlinear estimates are definitely
sharper in appearance when compared to the linear estimates computed using the Gauss-Markov
random field model in Figure 9.14. ( Continued )

of 3.5 MB/s), degradation of DV frames is barely noticeable. 28 Superresolution
video enhancement was applied directly to a short sequence acquired by
panning the MiniDV camera over a stationary scene containing a clock, with
the enhancement results shown in Figure 9.18. DV frames were imported via