Book Image

Hands-On GPU-Accelerated Computer Vision with OpenCV and CUDA

By : Bhaumik Vaidya
Book Image

Hands-On GPU-Accelerated Computer Vision with OpenCV and CUDA

By: Bhaumik Vaidya

Overview of this book

Computer vision has been revolutionizing a wide range of industries, and OpenCV is the most widely chosen tool for computer vision with its ability to work in multiple programming languages. Nowadays, in computer vision, there is a need to process large images in real time, which is difficult to handle for OpenCV on its own. This is where CUDA comes into the picture, allowing OpenCV to leverage powerful NVDIA GPUs. This book provides a detailed overview of integrating OpenCV with CUDA for practical applications. To start with, you’ll understand GPU programming with CUDA, an essential aspect for computer vision developers who have never worked with GPUs. You’ll then move on to exploring OpenCV acceleration with GPUs and CUDA by walking through some practical examples. Once you have got to grips with the core concepts, you’ll familiarize yourself with deploying OpenCV applications on NVIDIA Jetson TX1, which is popular for computer vision and deep learning applications. The last chapters of the book explain PyCUDA, a Python library that leverages the power of CUDA and GPUs for accelerations and can be used by computer vision developers who use OpenCV with Python. By the end of this book, you’ll have enhanced computer vision applications with the help of this book's hands-on approach.
Table of Contents (15 chapters)

Histogram calculation in PyCUDA

The histogram of an image conveys important information related to the contrast of an image, and it can also be used as an image feature for computer vision tasks. A histogram indicates the frequency of the occurrence of a particular pixel value. While calculating the histogram of an 8-bit image that is 256 x 256 in size, the 65,535-pixel values will work on arrays of intensity values from 0-255. If one thread is launched per pixel, then 65,535 threads will work on 256 memory locations of intensity values.

Consider a situation in which a large number of threads try to modify a small portion of memory. While calculating the histogram of an image, read-modify-write operations have to be performed for all memory locations. This operation is d_out[i] ++, where first d_out[i] is read from memory, then incremented, and then written back to the memory...