受新兴的扩散模型（Diffusion Model）启发，
思考到传统CPU滤波算法无法适应大模型时代的大数据量。
考虑到使用并行计算的方法进一步提高算力，
以训练更大的模型，形成新的商业模式。
图像滤波器是为了解决噪点问题而建立的一个数学模型，通过这个模型来将图像数据进行能量转化。本项目基于C++/SYCL工具，实现对均值滤波与高斯滤波并行加速。
包括并行计算下的高斯滤波实现，积分图优化的均值滤波
2023/07/15🤗🤗 We, as the participating project in the "2023 Spring Intel oneAPI Campus Hackathon Competition" won the third prize and a special award, and received a 10,000 yuan reward from Intel Corporation. Thank you for Professor Dai Hongju's guidance.
| Optimized for | Description |
|---|---|
| OS | Ubuntu* 18.04 |
| Software | Intel® oneAPI DPC++/C++ Compiler |
| Dependency | OpenCV |
cmake .
make
./main {File_Name}There are three sample picture in the folders. They are "gaussian.png", "images.jpg" and "lena.png".
You can use these command to take a try.
./main gaussian.png
./main images.jpg
./main lena.png
The program will build three picture "gaussianOut.jpg","meanOut.jpg" and "meanPlusOut.jpg".
you need to apply for Computing nodes with GPU.
Using the follow command, you will get into the interactive mode.
qsub -I -l nodes=1:gpu:ppn=2 -d .
Then, you can use the same command On Linux to execute the program.
You can see the detail how the program implements in the file "文档.md" in the repository.