Acceleration of hough transform algorithm using Graphics Processing Unit (GPU)

The Hough Transform is a popular tool for feature detection in digital image processing due to its robustness to noise and missing data. However, the computational cost associated to its voting scheme has prevented software implementation to achieve real time performance. This paper demonstrates the parallel computing power of Graphics Processing Unit (GPU) in area of image processing. In this paper, we have presented an algorithm for performing Hough Transform on digital images using NVIDIA GPU with the support of MATLAB Parallel Computing Toolbox in order to achieve faster execution. The advantage of using GPU for general purpose computation is to speedup the performance which can be achieved by the parallel architecture of GPU. In this paper, we have harnessed this property of GPU to accelerate the process of Hough Transform for shape or feature detection and the performance of processing of Hough Transform using CPU and GPU are compared.

I. Introduction

The use of computer vision in real time applications is increasing with increase in use of mobile devices with embedded cameras and image processing techniques. Some of the applications of computer vision are autonomous vehicles, satellite image processing, medical image processing, traffic surveillance, consumer applications like the augmented reality on mobile phones. In the field of automated digital image analysis, shape detection is challenging task and is important step in object recognition. The manual extraction of shape information from an image can be very time consuming especially if there are more number of required shapes in the image. In such applications, an automatic method for shape detection is preferable.