Skip to main content
EURASIP Journal on Image and Video Processing
Figure 3 | EURASIP Journal on Image and Video Processing

Figure 3

From: Handling missing weak classifiers in boosted cascade: application to multiview and occluded face detection

Figure 3

Differences between a cascade and a McCascade for the classification of a sample x. In a cascade, all weak classifiers are used. At level j, H j (x) is computed and is compared to the threshold τ j . An occluded face is most of the time rejected because occluded subwindows corrupt the decision on each level. In a McCascade, only a subset of weak classifiers is used. In the figure, only weak classifiers in charge of classifying the upper part of the face are used. At level j, P(y 1:j =1|x) is computed and is compared to the threshold β j . An occluded face is most of the time detected. In contrast to the decision at level j of a cascade, the decision at level j of a McCascade incorporates the decision of previous levels.

Back to article page