基于CS-BP神经网络的舌诊图像颜色校正算法

CS-BP Neural Network-based Color Correction Algorithm for Tongue Image

针对开放环境下舌诊图像采集过程中存在颜色偏差问题,文章提出基于CS(布谷鸟搜索)-BP的舌象颜色校正算法,利用布谷鸟的巢寄生性以及levy飞行机制优化BP神经网络。为了与其他颜色校正算法作对比,文章选择了多项式回归。为了研究不同拍摄环境对颜色校正结果的影响,分别在不同时刻下的室内、室外、白炽灯拍摄环境下,采集带有24色色卡的舌象并应用三种算法对其颜色校正得出结果进行对比分析,采用CIElab色差值指标对这三种算法进行评价,实验结果表明,与多项式回归和BP神经网络算法相比,CS-BP算法的校正效果得到明显提高。

The tongue color correction algorithm based on CS( cuckoo search)-BP was proposed to address color deviation in the tongue image acquisition process in an open environment. The BP neural network was optimized by using cuckoo’s nest parasitism and levy flight mechanism. The polynomial regression was used to make a comparison between the CS-BP based algorithm and other color correction algorithms. To study the influence of different shooting environments on the color correction,the tongue images with 24-color cards were taken indoors,outdoors and incandescent light respectively at different times. Three different algorithms were then applied to make a comparative analysis on the color correction. The CIElab color difference index was used to evaluate the three algorithms. The experimental results show that the correction effect is greatest when using the CS-BP algorithm compared with the polynomial regression and the BP neural network algorithm.