基于Bubble小波的多尺度边缘提取

Multiscales Edges Detection Based On Bubble Wavelet

侧抑制是生物视觉信息处理中广泛存在的一种机制，Ｂｕｂｂｌｅ函数很好刻画了侧抑制现象。本文利用Ｂｕｂｂｌｅ函数构成小波，高斯函数作为平滑函数，用Ｍａｌｌａｔ快速算法进行小波分解，分解的局部极大值就是多尺度边缘，理论和实验都表明，这种算法对于不同的尺度，有不同的抑制噪声的能力和提取边缘细节的能力，当Ｂｕｂｂｌｅ函数的尺度较大时，抑制噪声的能力增强，提取边缘细节的能力变差；当Ｂｕｂｂｌｅ函数的尺度较小时，抑制噪声的能力变弱。

Lateral inhibition described by Bubble function is a very common mechanism in living creature vision information processing system.In this paper,we use Bubble function as wavelet base and Gaussian function as smooth function to decompose signal using Mallat algorithm and the position of maximum decomposition value is just the position of edges.This algorithm has different ability to reduce noise and to extract edges according to different scales of Bubble function.When the scale of Bubble function is growing wide,then the ability of reducing noise increases and the ability of extracting detail edges decreases;otherwise,the ability of reducing noise decreases and the ability of extracting detail edge increases. Both the theory and experiment manifest that this algorithm is very efficient and practical.