Performance comparison of OpenCV applications with and without CUDA support

The performance of image processing algorithms can be measured in terms of time it takes to process a single image. When algorithms work on video then performance is measured in terms of frames per second which indicates the number of frames it can process in a second. When algorithm can process more than 30 frames per second then it can be considered to work in real time. We can also measure the performance of our algorithms implemented in OpenCV which will be discussed in this section.

As we have discussed earlier, When OpenCV is built with CUDA compatibility it can increase the performance of algorithms drastically. OpenCV functions in CUDA module are optimized to utilize GPU parallel processing capability. OpenCV also provides similar functions that only run on CPU. In this section, we will compare the performance of thresholding operations built in the last section with and without using GPU. We will compare...