How to use WarpedView to interpolate a 3D image

Specific Domains
, , ,
1

I’m referring to the WarpedView function from ImageTransformations.jl: ImageTransformations.jl · ImageTransformations. I can use imresize to interpolate a 3D image naively like: imresize(image, ratio=(1,1,z_res/xy_res)), where z_res/xy_res is the ratio of the z to the xy resolution. I think WarpedView is the appropriate function to do this operation lazily but I cannot tell from the documentation how to actually do it. Anyone know how to do so? Thanks.

2

imresize defined by the given ratio can be defined as an image warping by the linear transformation of matrix:

[ 1 0 0 ; 
  0 1 0; 
  0 0 z_res/xy_res]

But the image warping is implemented via backward transformation:

using Images, ImageTransformations, CoordinateTransformations, Interpolations
function  invtform(xy_res, z_res)  #inverse transformation
    M=[1 0 0 ; 0 1 0; 0 0 xy_res/z_res]
    LinearMap(M)
end 
img = load("image-3d.tif");
xy_res = 120
a, b, c= size(img)
imgw =  warpedview(img, invtform(xy_res, c), 
          (1:a, 1:b, 1:round(Int, c^2/xy_res)); method=Lanczos(4))
1 Like
3

Hi, thanks for the response. I should have clarified: what I meant by xy_res and z_res was the physical resolution per pixel in each direction. So, for example, 1 micron as the xy_res and 4 microns as the z_res. In your usage, these refer to the height, width, and number of slices of the image.

4

I interpreted xy_res as DPI, and z_res as number of slices.You shoud derive the image transformation and its inverse in the case of the right interpretation of the two parameters.

1 Like
5

Okay, got it. Just wondering though – where did c^2/xy_res come from as the third index?

6

c is the last element in 1:c. This value multiplied by z_res/xy_res, and rounded, gives the last value of the third index in the warped image. I stress again, the image transformation/warping is performed by the initial linear map. Its inverse is used only in the implementation of warpedview.

2 Likes