我正在使用 SimpleElastix ( https://simpleelastix.github.io/ ) 进行两个 2D 图像的注册 (仿射) (见附件) 。为此,我正在使用此代码:
import SimpleITK as sitk
elastixImageFilter = sitk.ElastixImageFilter()
elastixImageFilter.SetFixedImage(sitk.ReadImage("fixed_image.nii"))
elastixImageFilter.SetMovingImage(sitk.ReadImage("float_image.nii"))
elastixImageFilter.SetParameterMap(sitk.GetDefaultParameterMap("affine"))
resultImage=elastixImageFilter.Execute()
sitk.WriteImage(resultImage,"registred_affine.nii")
后者执行后,我获得以下包含变换矩阵的 TransformParameters0.txt :
(Transform "AffineTransform")
(NumberOfParameters 6)
(TransformParameters 0.820320 0.144798 -0.144657 0.820386 -13.106613 -11.900934)
(InitialTransformParametersFileName "NoInitialTransform")
(UseBinaryFormatForTransformationParameters "false")
(HowToCombineTransforms "Compose")
// Image specific
(FixedImageDimension 2)
(MovingImageDimension 2)
(FixedInternalImagePixelType "float")
(MovingInternalImagePixelType "float")
(Size 221 257)
(Index 0 0)
(Spacing 1.0000000000 1.0000000000)
(Origin 0.0000000000 0.0000000000)
(Direction 1.0000000000 0.0000000000 0.0000000000 1.0000000000)
(UseDirectionCosines "true")
// AdvancedAffineTransform specific
(CenterOfRotationPoint 110.0000000000 128.0000000000)
// ResampleInterpolator specific
(ResampleInterpolator "FinalBSplineInterpolator")
(FinalBSplineInterpolationOrder 3)
// Resampler specific
(Resampler "DefaultResampler")
(DefaultPixelValue 0.000000)
(ResultImageFormat "nii")
(ResultImagePixelType "float")
(CompressResultImage "false")
我的目标是使用这种矩阵变换来注册浮动图像并获得与 SimpleElastix 获得的类似的注册图像。为此,我正在使用这个小脚本:
import SimpleITK as sitk
import numpy as np
T= np.array([[0.82, 0.144, -13.1], [-0.144, 0.82, -11.9], [0, 0, 1]] ) #matrix transformation
img_moved_orig = plt.imread('moved.png')
img_fixed_orig = plt.imread('fixed.png')
img_transformed = np.zeros((img_moved_orig.shape[0],img_moved_orig.shape[1]))
for i in range(img_moved_orig.shape[0]):
for j in range(img_moved_orig.shape[1]):
pixel_data = img_moved_orig[i, j]
input_coords = np.array([i, j,1])
i_out, j_out, _ = T @ input_coords
img_transformed[int(i_out), int(j_out)] = pixel_data
我获得了这个注册图像,我将它与 SimpleElastix 的结果进行了比较(见附图)。我们可以观察到缩放没有操作,平移有问题。我想知道我是否遗漏了转换矩阵中的某些内容,因为 SimpleElastix 提供了良好的配准结果。
有任何想法吗 ?
谢谢