tps.unwarp: Non-linear image registration using TPS deformation.

In J0vid/Colormesh: Functions for Processing Image Color Data for Statistical Analysis

Description

Non-linear image registration using TPS deformation.

Usage

tps.unwarp(
  imagedir,
  landmarks,
  image.names,
  write.dir = NULL,
  sliders = NULL,
  target = NULL
)

Arguments

imagedir	Directory of images to deform. Only images with landmarks will be processed. The landmark file names are assumed to exactly match the image names.
landmarks	A landmark array with dimensions N_landmarks x 2 x N_observations. dimnames(landmarks)[[3]] should have the corresponding image filenames for each observation.
image.names	A vector of image names to look for in imagedir. These images should be unwarped or deformed to a common reference shape.
write.dir	Where to save warped images. Images will be named after the original image name (is that a bad idea because of overwriting? We will find out).
sliders	An index of sliding semilandmarks for calculating the mean landmark shape.
target	Supply a target shape to warp to. The default is the mean shape. The target will be placed in the center of the image for unwarping.

Value

warped images will be saved to the write.dir directory. We also return the consensus shape of the landmarks. This can be used for delaunay triangulation. Finally, we return the unwarped image names.

Examples

#load landmarks and covariate data
guppy.lms <- tps2array(system.file("extdata", "original_lms.TPS", package = "Colormesh"))
specimen.factors <- read.csv(system.file("extdata", "specimen_factors.csv", package = "Colormesh"), header = F)

#define sliders for guppy data
sliders <- make.sliders(c(1,8:17, 2, 18:19, 3, 20:27, 4, 28:42, 5, 43:52, 6, 53:54, 7, 55:62), main.lms = 1:7)

#unwarp images--change writedir if you want to see the images!
example.sample <- tps.unwarp(imagedir = paste0(path.package("Colormesh"),"/extdata/cropped_images/"), landmarks = guppy.lms, image.names = specimen.factors[,1], sliders = sliders, write.dir = tempdir())

J0vid/Colormesh documentation built on Feb. 4, 2022, 1:33 a.m.