Image Processing Reference
coefficients in the outer periphery are replaced by a linear combination of the wave-
let coefficients in the inner periphery. The parameters of the combination are a
function of the distance to the edge. Finally, an inverse wavelet transform is applied
to produce the encrusted image with smooth edges.
In order to illustrate the application of the proposed method, it is applied to a test
case and compared to standard procedure (raw encrustation). In the test case, an
imagette is extracted from the reference image and is encrusted into a 5.6 m
smoothed image. Two encrusted images are made: one based on the raw procedure
(previously shown in Fig. 11.2 ), and one based on our proposed smooth encrusta-
tion procedure (shown in Fig. 11.4 below). An urban area has been selected for this
test case because this is certainly the most difficult type of landscape to process
from a numerical point of view, and therefore it helps point out the drawbacks and
qualities of algorithms. In addition, urban areas entail high variability of information
induced by the diversity of the features sizes.
In the raw encrusted image (Fig. 11.2 ), pixels have either an effective resolution
of 5.6 m (outside the imagette area) and 0.7 m (inside the imagette area). Large size
features such as streets can be followed across the image whatever the resolution.
Fig. 11.4 Low resolution image with a high-resolution encrusted imagette with improved outer