qianxunzdl123 / cuda-image-enhancement

An implementation of parallel version of gamma correction, histogram equalization, gaussian filter and sobel filter algorithms written in CUDA C++.OpenCV is used for including image into main program and saving output image.

An implementation of parallel version of gamma correction, histogram equalization, gaussian filter and sobel filter algorithms written in CUDA C++.OpenCV is used for including image into main program and saving output image.

GPU Specs : GTX 1660 TI Mobile (TU116 Core), 1536 CUDA cores, 24 SM.
CPU Specs : i7-9750H 2.4GHz base clock.

Speedups obtained by averaging 100 times execution on 1024x1024 image.
gamma correction : 2.29579 times faster

histogram equalization : 2.06096 times faster
histogram equalization (shared memory) : 2.53584 times faster

gaussian filter : 59.6874 times faster
gaussian filter (shared memory): 49.5195 times faster (bank conflict,gonna fix later)

sobel filter : 89.1415 times faster
sobel filter (shared memory) : 76.5014 (bank conflict, gonna fix later)

This application is currently in develop and maintenance stage.