Parallelization of template matching
- 0 Collaborators
The implementation of the parallel template matching algorithm based on oneapi reads the template image and the source image, slides the template image on the source image and calculates the difference between the corresponding image pixels, and finally returns the position with the minimum differen ...learn more
Project status: Published/In Market
Overview / Usage
Parallel template matching algorithm solves the problem of long running time of template matching algorithm. At the same time, the parallel template matching algorithm based on oneapi can be applied to multiple platforms, increasing the application scenarios of template matching algorithm. Usage method: input source image and template matching, you can find the closest position in the source image and template image.
Methodology / Approach
Problem solving process: first learn the oneapi document, understand the usage of related functions under the oneapi framework, and learn the related content of heterogeneous computing, and apply it in parallel on the template matching algorithm.
Framework: oneapi, which aims to provide a unified programming model and API interface for all kinds of computing architectures, so that programmers can write code only once, and then programs can be executed on CPU, GPU, FPGA and other hardware platforms.
Technologies Used
oneAPI: in this framework, the template matching algorithm is programmed in parallel, which can be used on multiple platforms
STB: library function for opening and saving pictures
Sycl: sycl supports single source code development, in which C + + template functions can contain host code and device code to build complex algorithms accelerated by OpenCL, and then reuse them in the source code of different types of data.