Image Processing Reference
In-Depth Information
Figure 4.3
Table of observations showing filter functions, error for the optimum filter, and the
earlier filter designed on purely additive noise.
remaining error is interesting. The filter trained on the earlier image with only addi-
tive noise matches the optimum filter in removing the noisy black pixels in the
background. It is, however, unable to repair any of the noisy white pixels on the
black text. It has also allowed the edges of the text to be thinned. The difference be-
tween the two images is shown in Fig. 4.2. The observation table and comparison
between the filters is given in Fig. 4.3. The errors may be easily computed as in the
previous chapters.
4.1 Effect of Window Size
As seen in the examples of the previous chapter, the error resulting from filtering
with the optimum filter in any given window reduces with an increase in window
