摘要
针对传统的图像重构算法的不足,提出一种基于小波神经网络的图像重建快速学习算法。运用小波神经网络对图像重构进行了仿真研究。实验表明,对于不同的误差模型,小波神经网络采用不同的基函数可以很好地对非线性系统进行逼近,收敛速度快,近似精度高,而且网络规模比较小,计算量少。对计算机视觉和图像处理具有良好的应用价值。
The limitation of the conventional Lambertian reflectance model for the image reconstruction is addressed and a new wavelet neural network （WNN）-based image reconstruction model is proposed. The new neural learning algorithm is to optimize a proper reconstruction model and to recover the object surface by a simple shape- form shading （SFS） variation method with this WNN-based model and fuzzy model. An example is also given to prove that the SFS technique is robust to most objects, even when the lighting conditions are uncertain. The simulation result shows the training speed of WNN can be improved greatly, and the method can be applied widely.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2005年第12期1985-1987,共3页
Systems Engineering and Electronics
基金
国家自然科学基金(60373062)
湖南省自然科学基金(04JJ3052)资助课题
关键词
小波神经网络
图像重构
训练算法
优化
wavelet neural network
image reconstruction
learning algorithm
optimization
