Image Processing Reference
In-Depth Information
Figure 9.14
The result achieved by using the SSIM-based training algorithm. Minor artifacts
can be seen, but otherwise the image is of a much higher subjective quality than the previous
figure.
It is not possible to quote values of the quality measure for these images since they
both require the original clean ideal images, which of course we do not have.
9.4 Hardware Implementation
The major processing step in the implementation of the soft morphological filter
(SMF) is the rank ordering of data. For single images such as those shown in this
chapter, the processing may be carried comfortably in software. The C++ program
to process the 1024 × 1024 astronomical images in this chapter takes approximately
100 milliseconds per image to complete on a Pentium 4 running at 1.8 GHz with
