问题描述
我正在使用包含感兴趣区域的轮廓坐标的 xml 文件处理医学图像。我能够提取这些点，但无法将其转换为可用于为医学图像创建遮罩的边界框。
以下是一个感兴趣区域的“point_px”。 'point_mm' 是一个类似的元组列表，全部为零。
[
(2462.929932,995.062988),(2452.830078,986.978027),(2442.719971,982.935974),(2430.914551,983.919678),(2420.48999,988.999023),(2409.370117,997.083984),(2402.300049,1008.200012),(2397.25,1019.320007),(2393.172363,1030.600708),(2392.179199,1041.526123),1052.451416),(2394.165527,1065.363159),(2399.131592,1079.268066),(2410.056885,1089.200195),(2421.975586,1094.16626),(2432.900879,1097.145996),(2443.659912,1095.849976),(2453.399902,1094.329956),(2465.969971,1090.060059),(2477.080078,1081.969971),(2485.169922,1071.869995),(2491.22998,1060.75),1049.630005),(2490.219971,1038.52002),(2484.159912,1025.380005),(2478.090088,1015.27002),(2472.030029,1004.159973)
]
图像上没有轮廓，因此我无法使用 findcontour 方法。
解决方法
您可以通过查找最低的 x,y 值和最高的 x,y 值（分别为左上角和右下角）来获得边界框。
import numpy as np
# given points
points = [
(2462.929932,995.062988),(2452.830078,986.978027),(2442.719971,982.935974),(2430.914551,983.919678),(2420.48999,988.999023),(2409.370117,997.083984),(2402.300049,1008.200012),(2397.25,1019.320007),(2393.172363,1030.600708),(2392.179199,1041.526123),1052.451416),(2394.165527,1065.363159),(2399.131592,1079.268066),(2410.056885,1089.200195),(2421.975586,1094.16626),(2432.900879,1097.145996),(2443.659912,1095.849976),(2453.399902,1094.329956),(2465.969971,1090.060059),(2477.080078,1081.969971),(2485.169922,1071.869995),(2491.22998,1060.75),1049.630005),(2490.219971,1038.52002),(2484.159912,1025.380005),(2478.090088,1015.27002),(2472.030029,1004.159973)]
# to numpy array
numped = np.array(points);
# get corners (top-left,bottom-right)
tl = [min(numped[:,0]),min(numped[:,1])];
br = [max(numped[:,max(numped[:,1])];
# print
print(tl);
print(br);
numpy slice numped[:,0] 将所有 x 值作为单个 1D numpy 数组获取
另一个切片 numped[:,1] 获取所有 y 值。