基于改进区域项CV模型的金相图像分割 
下载： 694次
Segmentation of Metallographic Image Based on Improved CV Model Integrated with Local Fitting Term
1 南京航空航天大学电子信息工程学院, 江苏 南京 211106
2 北京科技大学新金属材料国家重点实验室, 北京 100083
图 & 表
图 1. 不同方法对金相图像1的分割结果。(a)金相图像1; (b)传统CV模型; (c)测地线活动轮廓模型; (d)偏置场修正水平集模型; (e)局部二值拟合模型; (f)本文模型
Fig. 1. Segmentation results of metallographic image 1 by different methods. (a) Metallographic image 1; (b) traditional CV model; (c) geodesic active contours model; (d) bias field correction level set model; (e) local binary fitting model; (f) proposed model
下载图片 查看原文
图 2. 不同方法对金相图像2的分割结果。(a)金相图像2; (b)传统CV模型; (c)测地线活动轮廓模型; (d)偏置场修正水平集模型; (e)局部二值拟合模型; (f)本文模型
Fig. 2. Segmentation results of metallographic image 2 by different methods. (a) Metallographic image 2; (b) traditional CV model; (c) geodesic active contours model; (d) bias field correction level set model; (e) local binary fitting model; (f) proposed model
下载图片 查看原文
图 3. 不同方法对金相图像3的分割结果。(a)金相图像3; (b)传统CV模型; (c)测地线活动轮廓模型; (d)偏置场修正水平集模型; (e)局部二值拟合模型; (f)本文模型
Fig. 3. Segmentation results of metallographic image 3 by different methods. (a) Metallographic image 3; (b) traditional CV model; (c) geodesic active contours model; (d) bias field correction level set model; (e) local binary fitting model; (f) proposed model
下载图片 查看原文
图 4. 不同方法对金相图像4的分割结果。(a)金相图像4; (b)传统CV模型; (c)测地线活动轮廓模型; (d)偏置场修正水平集模型; (e)局部二值拟合模型; (f)本文模型
Fig. 4. Segmentation results of metallographic image 4 by different methods. (a) Metallographic image 4; (b) traditional CV model; (c) geodesic active contours model; (d) bias field correction level set model; (e) local binary fitting model; (f) proposed model
下载图片 查看原文
表 15种分割方法的DSC值
Table1. DSC values of five segmentation methods
| Image | CV model | Geodesic active contour model | Bias field correction level set model | Local binary fitting energy model | Proposed model |
|---|
| Metallographic image 1 | 0.834 | 0.652 | 0.892 | 0.640 | 0.913 | | Metallographic image 2 | 0.864 | 0.849 | 0.715 | 0.709 | 0.927 | | Metallographic image 3 | 0.780 | 0.773 | 0.867 | 0.831 | 0.902 | | Metallographic image 4 | 0.782 | 0.675 | 0.869 | 0.669 | 0.881 |
|
查看原文
表 25种分割方法的运行时间
Table2. Running time of five segmentation methodss
| Image | CV model | Geodesic active contour model | Bias field correction level set model | Local binary fitting energy model | Proposedmodel |
|---|
| Metallographic image 1 | 230.54 | 125.88 | 450.94 | 195.23 | 189.06 | | Metallographic image 2 | 235.80 | 129.02 | 473.82 | 202.02 | 190.94 | | Metallographic image 3 | 106.63 | 59.13 | 198.39 | 93.33 | 89.83 | | Metallographic image 4 | 323.97 | 167.17 | 653.21 | 289.06 | 275.83 |
|
查看原文
表 3Otsu阈值选择准则与倒数交叉熵阈值选取准则的阈值
Table3. Thresholds obtained by Otsu algorithm and reciprocal cross entropy algorithm
| Algorithm | Metallographic image 1 | Metallographic image 2 | Metallographic image 3 | Metallographic image 4 |
|---|
| Otsu | 109 | 139 | 137 | 157 | | Reciprocal cross entropy | 97 | 157 | 152 | 124 |
|
查看原文
倪康, 吴一全, 庚嵩. 基于改进区域项CV模型的金相图像分割[J]. 光学学报, 2018, 38(4): 0411009. Kang Ni, Yiquan Wu, Song Geng. Segmentation of Metallographic Image Based on Improved CV Model Integrated with Local Fitting Term[J]. Acta Optica Sinica, 2018, 38(4): 0411009.
PDF全文
