Meta-learning for Image Processing Based on Case-Based Reasoning - Computational Intelligence in Healthcare: Advanced Methodologies - page 249

Fig. 38. Dendrogram for CBR based on central moments (on binary gradient image)

Problematic is only neu3, which appears in the same group as neu4. Fig. 39 shows
in (b) the best possible segmentation result and in (c) the segmentation result with
the parameter (0.75, 2, 1). Fig. 39(c) is more oversegmented than Fig. 39 (b). To
get a better result, we can use a threshold for small edges (compare with section 5 and
Fig. 20).

(2,1,1) BS=98 (0.75,2,1) BS=180

(c) segmentation with
parameters of neu4

(a) Image neu3 (b) best segmentation

Fig. 39. The influence of the selected parameter-set on the segmentation result and the number
of basins shown in image neu3