invariantImageSimilarity: similarity metrics between two images as a function of...

In neuroconductor-devel/ANTsR: ANTs in R: Quantification Tools for Biomedical Images

Description

compute similarity metric between two images as image is rotated about its center w/or w/o optimization

Usage

invariantImageSimilarity(
  in_image1,
  in_image2,
  localSearchIterations = 0,
  metric = "MI",
  thetas = seq(from = 0, to = 360, length.out = 5),
  thetas2 = seq(from = 0, to = 360, length.out = 5),
  thetas3 = seq(from = 0, to = 360, length.out = 5),
  scaleImage = 1,
  doReflection = 0,
  txfn = NA,
  transform = c("Affine", "Similarity", "Rigid"),
  mask
)

Arguments

in_image1	reference image
in_image2	moving image
localSearchIterations	integer controlling local search in multistart
metric	which metric MI or GC (string)
thetas	numeric vector of search angles in degrees
thetas2	numeric vector of search angles in degrees around principal axis 2 (3D)
thetas3	numeric vector of search angles in degrees around principal axis 3 (3D)
scaleImage	global scale
doReflection	reflect image about principal axis
txfn	if present, write optimal tx to .mat file
transform	Rigid, Similarity or Affine transform
mask	optional fixed image mask

Value

dataframe with metric values and transformation parameters

Author(s)

Brian B. Avants

Examples

fi<-antsImageRead( getANTsRData("r16") )
mi<-antsImageRead( getANTsRData("r64") )
mival<-invariantImageSimilarity( fi, mi, thetas = c(0,10,20) )
mapped = antsApplyTransforms( fi, mi, transformlist=mival[[2]] )
areg = antsRegistration( fi, mi, typeofTransform="Affine",
  initialTransform=mival[[2]] )
bestInd = which.min( mival[[1]]$MetricValue )
txparams = as.numeric( mival[[1]][ bestInd,2:(ncol( mival[[1]] )-2) ] )
txfixedparams = as.numeric( mival[[1]][ bestInd,
(ncol( mival[[1]] )-2+1):ncol( mival[[1]] )] )
affTx = createAntsrTransform( type = "AffineTransform", dimension = 2,
  parameters = txparams, fixed.parameters = txfixedparams )
mapped2 = applyAntsrTransformToImage( affTx, mi, fi )

scaleMat = diag( 2 ) * 0.75
affTx = createAntsrTransform( type = "AffineTransform", dimension = 2,
     matrix = scaleMat, fixed.parameters = c(125.2706, 129.2100) )
temp = applyAntsrTransformToImage( affTx, mi, mi )
mival<-invariantImageSimilarity( fi, temp, thetas = c(0,10,20),
  localSearchIterations = 10, transform='Similarity'  )
mapped = antsApplyTransforms( fi, temp, transformlist=mival[[2]] )
mival<-invariantImageSimilarity( fi, temp, thetas = c(0,10,20),
  localSearchIterations = 10, transform='Affine'  )
mapped2 = antsApplyTransforms( fi, temp, transformlist=mival[[2]] )
print( cor( fi[ fi > 0 ], temp[fi>0] ))
print( cor( fi[ fi > 0 ], mapped[fi>0] ))
print( cor( fi[ fi > 0 ], mapped2[fi>0] ))

cos45 = cos(pi*45/180)
sin45 = sin(pi*45/180)
txRotate <- createAntsrTransform( precision="float", type="AffineTransform", dim=2 )
setAntsrTransformParameters(txRotate, c(cos45,-sin45,sin45,cos45,0,0) )
setAntsrTransformFixedParameters(txRotate, c(128,128))
rotatedImage = applyAntsrTransform(txRotate, mi, mi)
mival<-invariantImageSimilarity( fi, rotatedImage, thetas = c(0,10,20,30,40,50),
  localSearchIterations = 10, transform='Rigid'  )
mapped2 = antsApplyTransforms( fi, rotatedImage, transformlist=mival[[2]] )

neuroconductor-devel/ANTsR documentation built on April 1, 2021, 1:02 p.m.