摘要
核Adaline是在最小均方误差基础上,通过迭代产生回归函数,逼近目标函数,方法简单,速度快.引入支持向量机中的不敏感带,推广了核Adaline算法,并将其应用于图像去噪.实验证明,不仅可以有效去除尖峰噪声,而且对随机噪声也具有一定的抑制作用.
Kernel_based Adaline learning algorithm, presented on the basis of least mean square error, approximates object function by regression with iteration. This technique is easy to understand and runs faster to optimal solution. By introducing ε-insensitive zone of support vector machine, Kernel_Based Adaline algorithm is modified to get better performance. Image denoising as an application of the amendatory algorithm. It can not only effectively eliminate spike noise as outliers, but also smooth random noise within this ε-insensitive zone. Finally, analysis and experiments on the benchmark image and pretreatment image of photo show the superiority of this method.
