Bayesian Image and Video Enhancement Using a Non-Gaussian Prior - Nonlinear Signal and Image Processing: Theory, Methods and Applications - page 326

(a) (b)

(c) (d)

FIGURE 9.16
Bayesian superresolution video enhancement of the IKONOS satellite picture with an effective
GSD of 8 m. (a) Microview of the original 1-m image data. (b) 8-m image data. (c) Bayesian
enhancement of the 8-m pixels using 1 frame (GMRF). (d) Bayesian enhancement of the 8-m
pixels using 1 frame (HMRF, T H = 1 . 5). (e) Bayesian enhancement of the 8-m pixels using all

λ ( i ) = 4 available frames (GMRF, 1000). (f) Bayesian enhancement of the 8-m pixels using all 4

λ ( i ) = available frames (HMRF, T H = 1000). Note that the nonlinear (HMRF) estimates

are slightly sharper in appearance when compared to the linear (GMRF) estimates. ( Continued )

1 . 5 and

extraction of additional details from arbitrary MPEG sequences is generally
not possible. Examining Figure 9.18 reveals evidence that enough redundant
pixel information remains within an intra-frame coded DV sequence (low
fixed compression ratio of 5:1) to warrant the application of superresolution
enhancement to video coded at high bit rates.