Graphics Reference
In-Depth Information
Figure
.
.
Reconstructions of the noisy image from Fig.
.
. Upper row: nonadaptive smoothing;
central row: structurally adaptive reconstructions; bottom row: pointwise sum of weights used in the
structurally adaptive reconstructions. Let column: local constant smoothing, e.g., kernel smoothing and
AWS; central column: local linear models; right column: local quadratic models. All reconstructions use
MAE optimal values for the bandwidth or maximal bandwidth, respectively
and the MSE for the reconstruction are significantly smaller than those obtained for
local constant AWS, see Table
.
.
hedetailed viewofferedbyFig.
.
allowsforamoreprecisejudgmentoftheim-
age quality for one of the most structured regions in the image. We provide the same
segment of size
pixels of the original image, its noisy version, and both the
local constant andlocal quadratic reconstructions. helocal constant reconstruction
generally provides morecontrast, at the cost of introducing artifacts into the smooth
regions, such as the sculpture. Local quadratic PS gives a better result with respect to
optical impression, MAE and MSE.
