Indexing, Object Segmentation, and Event Detection in News and Sports Videos - Multimedia Database Retrieval: Technology and Applications - page 195

Fig. 7.17 Example images and the application of nonlinear histogram equalization: ( first column )
the original images; ( second column ) histograms of the original images; ( third column ) filtered
images; ( forth column ) histograms of the filtered images

Table 7.2 Face detection results for the two testing databases

Good Bad

illumination illumination Changing head

case (348 case (427 poise/sizes (249 Multiple faces

Methods samples) samples) samples) (143 samples)

GW 87.50 75.25 72.49 70.14

GW + LN 93.74 85.69 89.06 81.38

bad conditions, changing head poses/sizes and multiple faces, respectively. It can be
observed that the application of local normalization to the face detection algorithm
provides significant improvement in detection rates for all cases discussed.