Research Article

ROC-Boosting: A Feature Selection Method for Health Identification Using Tongue Image

Figure 4

Superposition of all selected features is shown. From (a) to (d):(a) is in grayscale and is used as the background, (b) is the result of Test 1, (c) is the result of Test 2, and (d) is the result of Test 3. In (b, c, d), every rectangle represents one feature. The lighter partition in the rectangle is the center of the feature, and darker partitions surround this feature. Squares at the four corners of the feature were not removed for simplification. When repeatedly overlaid while keeping the color of the darker partitions fixed, the lighter partition would continue to be lighter. Finally, the lightest partition is observed intensively.
(a)
(b)
(c)
(d)