https://github.com/sebftw/interp2gpu

GPU-accelerated 2D spline interpolation, à la interp2(..., "spline"), in MATLAB.
https://github.com/sebftw/interp2gpu

cuda gpu gpu-acceleration matlab spline spline-interpolation

Last synced: 2 months ago
JSON representation

GPU-accelerated 2D spline interpolation, à la interp2(..., "spline"), in MATLAB.

• Host: GitHub
• URL: https://github.com/sebftw/interp2gpu
• Owner: sebftw
• License: mit
• Created: 2024-09-17T14:23:45.000Z (5 months ago)
• Default Branch: main
• Last Pushed: 2024-12-09T09:19:34.000Z (2 months ago)
• Last Synced: 2024-12-09T10:26:00.098Z (2 months ago)
• Topics: cuda, gpu, gpu-acceleration, matlab, spline, spline-interpolation
• Language: C
• Homepage:
• Size: 896 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Fast 2D Spline Interpolation

__interp2gpu__ is a drop-in replacement to `interp2` to perform fast spline interpolation on the GPU in MATLAB.

## How to use

Call it as `Vq = interp2gpu(V, Xq, Yq, "spline")`.  

For an explanation of inputs, see  [https://mathworks.com/help/matlab/ref/interp2.html](https://mathworks.com/help/matlab/ref/interp2.html) from Mathworks.  

An example of using interp2gpu is given in [example.m](example.m).

## Installation

You can download the code with [this link](https://github.com/sebftw/interp2gpu/archive/refs/heads/main.zip).  

Then run `addpath(CODEPATH)`, where CODEPATH is the path of the directory where interp2gpu is located.  

You can now use interp2gpu!

## Features

:heavy_check_mark: Multiple images can be processed with one call to interp2gpu.  

:heavy_check_mark: The method "spline_approx" performs fast approximated spline interpolation.  

:x: Does not currently support double-precision inputs.  

:x: Does not currently support arbitrary input pixel grids.  

You can add support for double-precision inputs and non-uniform pixel grids relatively easily - feel free to contribute.

## Files

- `example.m` is an example that shows how to use interp2gpu.

- `interp2gpu.m` is the main function.

- `gpuThomas2D.m` is used to compute the image derivatives required to evaluate the spline.

- `get_kernel.m` is used to cache CUDA kernels, so they do not have to be loaded every time.

- `Spline_paper.pdf` is a paper containing the theory of the method and its application in ultrasound imaging..

- `kernels/getInterpolation2D.cu` contains the CUDA C++ source code used to evaluate the spline.

- `kernels/compile_kernels.m` is used to compile the CUDA code.

- `kernels/getInterpolation2D.ptx` is the compiled CUDA kernel.