Nonlinear Predictive Coding for Images by Using Back Propagation Network Trained by Levenberg-Marquardt Algorithm

摘要：

A nonlinear predictive coder for images by using BPN-LM (back propagation network trained by Levenberg-Marquardt algorithm) is presented in this paper. LM algorithm overcomes the demerits of the conventional gradient descent algorithm, such as slow convergence and tendency to run into a local minimum. The input of BPN is obtained by a nonlinear transformation of the pixels within the predictive area. Therefore, the nonlinear correlation among pixels in an image is exploited. In addition, the number of hidden nodes of the network is designed to adjust adaptively, so the flexibility of applications is improved. Experiments show that BPN-LM can converge to the performance target rapidly and a nonlinear predictor by using a trained BPN can achieve better performance than a linear predictor

关键词： Nonlinear predictor BPN LM algorithm

作者: Xuedong Liu Jian Liu Yihua Tan

作者单位: the School of Information Engineering, Wuhan University of Technology, Wuhan, China the Institute for Pattern Recognition and Artificial Intelligence, Huazhong University of Science an

会议类型: 国际会议

会议名称: The Second International Symposium on Intelligence Computation and Applications(ISICA 2007)(第二届智能计算及其应用国际会议)

会议地点: 武汉

会议语种:英文

在线出版日期: 2007-09-21（万方平台首次上网日期，不代表论文的发表时间）