基于MRF融合的银桂花朵图像阈值分割方法研究

Research on O. fragrans Albus image thresholding segmentation method based on MRF fusion

图像分割是花卉类别图像识别过程中的重要步骤,分割结果的优劣直接影响识别结果的准确性。针对自然场景下的银桂花朵图像,提出一种基于马尔科夫随机场(Markov Random Field,MRF)的阈值分割融合图像分割算法。首先提取RGB彩色图像中的R通道、G通道、B通道的灰度图,用Otsu、Kittler、Niblack、Kapur四种阈值分割算法对灰度图进行二值化,然后利用像素局域空间能量与图像间能量建立MRF总能量泛函,最后对能量泛函进行最小化迭代求解,得到融合后的分割图。试验结果表明,提出的算法能降低树干背景影响,分割效果好,能很好地提取银桂花朵,SD、Dice、ER、NR平均值分别为93.07%、96.35%、7.73%、1.30%。

Image segmentation is an important step in the process of flower classification recognition.The quality of segmentation results directly affects the accuracy of recognition results.A thresholding segmentation and fusion algorithm based on Markov random field(MRF)is proposed for the image of O.fragrans Albus in natural scenes.Firstly,the grayscale images of R channel,G channel and B channel in RGB color image are extracted,and the grayscale images are binarized by Otsu,Kittler,Niblack and Kapur thresholding segmentation algorithms.Then the total energy functional of MRF is established by using the local space energy of pixels and the energy between images.Finally,the energy functional is solved iteratively by minimizing the energy function and segmented image is obtained.The experimental results show that the proposed algorithm can reduce the influence of the tree trunk background,and has good segmentation effect.The average SD,Dice,ER and NR values are 93.07%,96.35%,7.73%and 1.30%,respectively.