File Exchange

image thumbnail

2D interpolation

version 1.1 (3.74 KB) by

Fast 2D linear interpolation of scalar of vector valued 2D images.



View License

 ZI = mirt2D_mexinterp(Z,XI,YI) interpolates 2D image Z at the points with coordinates XI,YI. Z is assumed to be defined at regular spaced points 1:N, 1:M, where [M,N]=size(Z).
If XI,YI values are outside the image boundaries, put NaNs in ZI.

The function is similar to Matlab's native
ZI = INTERP2(Z,XI,YI,'linear',NaN),
 but is much master. If Z is vector valued 2D image (3D array), then iteratively interpolates Z(:,:,1), Z(:,:,2),Z(:,:,3),.. etc. This works faster than to interpolate each image independently, such as
 ZI(:,:,1) = INTERP2(Z(:,:,1),XI,YI);
 ZI(:,:,2) = INTERP2(Z(:,:,2),XI,YI);
 ZI(:,:,3) = INTERP2(Z(:,:,3),XI,YI);
The extra speed gain is from the precomputation of coefficients for interpolation, which are the same for all images. Interpolation of a set of images is useful, e.g. for image registration, when one has to interpolate image and its gradients or for vector valued images such as RGB.

Limitations: only works for images on a regular grid, and only with linear interpolation method.

Don't forget to compile first:
mex mirt2D_mexinterp.cpp

Enjoy! See more at my homepage:

Comments and Ratings (2)

Great job! (I couldn't get to work CUDA versions.) Thank you!

very useful for my applications, and very nice code! Thanks Andriy!



small fix

MATLAB Release
MATLAB 7.4 (R2007a)

Download apps, toolboxes, and other File Exchange content using Add-On Explorer in MATLAB.

» Watch video