基于增益控制Lucy-Richardson算法的脐橙模糊图像复原

Restoration of Blurred Citrus Image Based on Gain-Control Lucy-Richardson Algorithm

在利用机器视觉系统对脐橙进行变袋长包装的过程中,脐橙的匀速直线运动会造成图像模糊进而导致脐橙边缘细节等特征信息丢失,极大地影响了脐橙尺寸检测精度。针对该问题,本文提出了一种求解脐橙图像模糊长度及恢复模糊图像的方法。对退化图像进行傅里叶变换(fourier transforn,FT),估计退化图像点扩展函数(point spread function,PSF),利用傅里叶频谱特性对运动模糊长度进行了计算,并采用了基于增益控制的Lucy-Richardson改进算法对运动模糊图像进行了Matlab复原仿真实验,结果表明:当算法迭代次数为8时,去振铃效应(ringing artifacts,RA)最明显,其图像分割误差最小。综合运用RGB颜色分量线性运算对传送链上的脐橙进行背景分割,对图像进行去噪处理以及二值形态学运算,利用最小外接矩形法(minimum enclosing rectangle,MER)计算脐橙本体的最大横径。实验结果表明,相对于原始未处理的模糊图像,使用本文算法使脐橙最大横径平均测量误差从5.1%下降至0.81%,提高了脐橙的包装精度。

In the process of citrus packaging via machine vision system,linear motion blurs the image and results in the loss of edge details,which greatly affected the measurement accuracy.In view of the problem,a method for resolving the image blur length and restoring the blurred image is proposed in this article.Fourier transform was conducted on the degraded image and point spread function of the image was estimated.The motion blur length was calculated by analysing the characteristics of Fourier spectrum,and gain-based Lucy-Richardson algorithm was applied to restore the motion degraded image.The simulation results showed that blurred image could get its best restoration with minimum segemntation errors when the iteration number N was equal to 8,compared with the conventional Lucy-Richardson algorithm with 10 iterations.Integrating linear RGB color component operation into segmenting citrus from the background of conveyor chain and plates,de-noising and binary morphological operation were carried out before detecting citrus edge,the maximum transverse diameter of citrus was calculated by using maximum enclosing rectangle.The simulation results showed that compared with the original unprocessed blurred image,its average absolute demensional measurement error dropped from 1% to 0.81%,the citrus packaging accuracy was improved.