Redundancy Reduction in Face Detection of Viola-Jones using the Hill Climbing Algorithm

In this study, we improved Viola-Jones face detection using Hill Climbing algorithm in order to reduce the redundancy rates. Viola-Jones's algorithm has succeeded in determining the best face scale factor, but it produces a non-comparable face window that leads the redundant face detection. Hill Climbing algorithm which is one of the local search family, is proposed to serve the local-maxima which represents a set of faces that has been selected from redundant data. We have evaluated and compared the accurate, accurate and recall tests the performance of the proposed method using LFW dataset. Traditional Viola Jones achieves accuracy up to 77% and the proposed method achieves accuracy up to 85%. The two sets of values concluded that the proposed method reduce the redundancy problem.