Cascaded split-level colour Haar-like features for object detection

Rectangular features have been widely used in different cascade-based object detection areas. Previous rectangular features contain no relationship of different colour channels, which makes the previous rectangular features weak in expressing objects with colour contrast. To overcome this shortcoming, a series of rectangular features called split-level colour Haar-like (SC-Haar-like) features are proposed. There are different colour channels in the SC-Haar-like features, and the different colour channels have close relationship with each other based on the split-level structure. The experiments on face detection show that the SC-Haar-like features outperform previous rectangular features in terms of detection rate and false alarm rate.