基于生物地理学优化算法的多尺度Retinex权重选择

Weight Selection of Multi-scale Retinex Based on Biogeography Optimization Algorithm

研究了采用生物地理学优化算法(BBO)的多尺度Retinex权重选择仿生图像增强算法.原始的多尺度Retinex对不同尺度参数的高斯滤波器进行权重平均分配,因而无法保证在多种应用场景下皆能获取最优的增强结果.为解决这一问题,采用BBO算法进行权重启发式搜索选择,对多尺度Retinex的结果加以优化.该方法能够弥补各个尺度高斯滤波器所占权重的缺陷.通过与原始多尺度Retinex增强的对比仿真结果,对算法的性能和灵活性进行了验证.

We present an implementation of weight selection in multi-scale retinex( MSR) using the biogeographybased optimization algorithm( BBO). The existing MSR algorithm provides the same values to the Gaussian filters with different scales and cannot ensure the best enhancement results in various environments. We use BBO for weight selection to optimize the enhancement in MSR. The optimization of weight selection compensates for the defects of the scale parameters in Gaussian filters. Extensive simulations with comparisons to the existing MSR verify that our method achieves better enhancement results.