LDA based Reduced Joint Integral Histogram for Feature Extraction Case of study: Face Detection

The face pattern is described by extracted features using the new Reduced Joint Integral Histogram (RJIH) data structure. Extending the classical representations of integral images and integral histograms, it joins the global information of two images. Then, we turn to Linear Discriminant Analysis (LDA) to project the obtained Joint Integral Histogram from d−dimensional subspace to one dimensional subspace. Best features are selected by Adaboost learning framework. The experimental results demonstrate that our proposed method RJIH under Adaboost training process further improve the performance of the JIH in terms of detection rate and false positive rate.