Visual Media Material Editing System Based On GPU

Visual media mainly contains images and videos, which are intuitional, vivid andinteresting, and can convey information to people more efectively compared with text.In the old days, traditional visual media such as painting and printing, and its carrierslike papers and flms couldn’t be in full use because of the high cost of creation andpreservation techniques and lack of efective reusing methods. However, nowadays thedevelopment of information technology and Internet changes the whole thing. On onehand, a variety of digital devices make it easier and cheaper to create and keep visualmedia, and on the other hand, the Internet takes full advantage of its power acting asmedium. Nowvisualmediahasbecomeoneofthemostimportantinformationcarriers,and we can see it everywhere in our daily life, and the need for it is still larger andlarger. As a result, it’s meaningful to study and develop the editing toolkit of the visualmedia. Because of the business competition, position-setting and so on, existing visualmedia editing tools can’t well satisfy all of editing, managing and system secondarydevelopment.It’s more and more difcult to improve as the Moore’s law the computational abil-ity of CPU, which is constrained by things like material, manufacturing technique andheat radiation with the development of computer hardware. People have changed theirminds to the direction that making multiple modules work in parallel. The powerfulcomputational performance in parallel of GPU has got people’s attention with the ar-chitecture of GPU maturing, and people start to let GPU do general computing besidesgraphics, and have achieved good results.This paper focus on the visual media material editing system based on GPU, inorder to implement a platform that has basic function of management and editing to thevisual media, and a friendly framework to the developers who can easily implementalgorithms on GPU about visual media based on this system. This paper introduces the relative background about visual media and parallel computing, especially analyzingthe problems of the existing tools and giving the importance of parallel computing.Then this paper introduces the techniques and background of general purpose on GPUin details from the history of GPU, architecture to its main idea. Non-photorealisticrendering, one of the most important areas of visual media editing, is also introduced.This paper gives a requirement analysis in details on the visual media material editingsystem based on GPU, give the design and implementation of each module of the sys-tem, and then give the design and implementation of the GPGPU framework. In thepart of system application, this paper takes the cartoon stylization as an example, andgives its algorithm and implementation on this system, and at last gives the result andcomparison to other methods. From the system design and the result, the visual mediamaterial editing system based on GPU implemented by this paper has basic function ofvisual media management and powerful functions of editing, and it’s also expandableand friendly to the developers, with a good prospect.