The Optimization Of FFmpeg Blur Filter Algorithm Based On OpenCL

FFmpeg, as a rapidly growing class of powerful audio and video processingsoftware, it can be used of recording and transforming both digital audio and video.At the same time, because of the cross-platform, free and the open source code.FFmpeg was developed under Linux environment firstly, however, in order to satisfythe market and the customer demand, through the cross platform technology FFmpegcan be employed in Windows and other environments. In addition, it also includesnumerous leading audio and the video codec library.However, with the rapid development of information age, more and more videoimage frames need to be processed, and the requirement of processing speed becomesmore and more strict, the goal that people pursuit of is high efficiency and short time.As a result, the traditional processing method that based on CPU processing of thefilters in FFmpeg seems overburdened, and shows up the problem of low efficiency ofprocessing.This article discussed how to solve the above problems. In this article, using thetypical video blur filters(unsharp filter) that from FFmpeg as the example, using GPUtechnology to optimize, using OpenCL technology to transform the code of the filtersfrom CPU to GPU, which can realize the effect that synchronous executemultithreading,so that achieve the objective of algorithm optimization.This article focuses on the research meaning and background firstly, which canlet the readers have a general understanding. Then, show the current research status,after analyzing, find out the shortcomings and the insufficiency of the presentresearch situation, and obtain the major research direction. Based on the above, thisarticle will show the related information of GPU structure and OpenCL to have a further understanding about this article. This article will mainly introduce and analyzethe Blur Filter Algorithm, pay more attention on Gaussian Blur. Learn the principlethrough the analysis of the structure of Gaussian function, the calculation of GaussianMatrix, the convolution between the picture and the Gaussian Matrix. On this basis,summarize the code that can be optimized though deep analyze, show the optimizedcode and the principle as well. Finally, the algorithm is evaluated by different testingmachine to analyze the results, and exhibit the corresponding forms and icons, whichcan be show the research degree of optimization.