Research And Implementation Of Gradient Field Image Enhancement Algorithms Based On GPU

Recently, with the advancement of science and technology, image enhancement has become an important research topic, has become an indispensable part of machine vision, has been widely applied in all fields in computer vision. Image enhancement has penetrated into all aspects of human life and social production, such as satellite navigation, multimedia communications, medical diagnostics, traffic detection fields, etc. In some cases, the visual effect of image is not ideal.The main purpose of image enhancement is to improve the visual effect of the image. In different applications, the information needed to be enhanced is different. Therefore, we need purposefully emphasizes whole or partial characteristics of the image, emphasize certain useful information, inhibit characteristic that is useless. Thus the image contains more abundant amount of information and the visual effect is better to meet the needs of some special analysis and applications.This paper studies the currently used image enhancement algorithms and analyzes their advantages and disadvantages, an improved gradient field enhancement algorithm is put forward according to the human visual system and actual requirements.The experiments show that not only does the proposed algorithm improve the image contrast, suppress the halo phenomenon but it is more in line with the human visual system and increase the proportion of pixels in the image which are greater than JND, which is suitable for the practical application of the image enhancement requirements.This paper describes the kernel function, the thread structure, the storage model, the relationship between the CPU and GPU and the programming model of CUDA. And analyzes the GPU hardware architecture and hardware mapping, and the launch and execution of warp. On this basis, expand research and implementation of GPU in image enhancement. This paper uses the architectural model of CPU plus GPU, CPU handles the serial work of strong logic while the GPU is responsible for a large amount of computation parallel work. They complete the image enhancement tasks together.The gradient field image enhancement algorithm which is based on GPU is realized utilizing CUDA parallel computing platform, C++ and OpenCV. In this paper based on the analysis of the original gradient field image enhancement algorithms, a parallel gradient field image enhancement algorithm which is applicable to CUDA platform is proposed. This paper presents a modular analysis of parallel algorithms and analyzes the algorithms through several aspects. Finally, the paper analyzes the advantages and disadvantages of parallel algorithms through experimental results and presents an optimization scheme, which shows that the proposed method in this paper improves the processing speed of about 30 times than traditional CPU-based enhancement algorithms.