Research On Multi-core DSP Platform-Based Image Deblurring Method

With the widespread use of machine vision,and portable imaging equipment is becoming more common.Image deblur technology is being used on a large scale and ushering in the round of rapid development.As a pre-processing method of machine vision detection and recognition technology,which can obtain more accurate and perfect effect.It is well known that during the imaging process of the portable camera,due to human factors,the system itself,the natural environment and many other factors,the exposure time of camera and the actual scene may have relative motion,and image blur is the camera sensor received exposure time accumulated light intensity.Image blur is the result of the camera photographic elements to receive accumulated light intensity in exposure time.However,it is difficult to extract effective target feature from the blurred image object region,which makes it difficult to guarantee the subsequent detection and recognition effect,reduce the recognition precision and even cause the failure of the whole recognition algorithm.Therefore,it becomes an indispensable part of machine vision technology,to use image deblur technology to restore the image of the target feature.Due to the complexity of the image deblur technology and the diversity of image blurring,the existing methods mainly adopt different methods for different degradation models.The deblurring results depend on the degradation pattern.Software complexity and parameter adjustment and more time-consuming,and the image recovery is poor.Therefore,a universal unified image denoising algorithm is studied in this paper.On the basis of studying its defuzzification principle,the core of high performance multi-core DSP is studied,the characteristics of embedded soft and hard platform are researched,and the image deblurring software system.Due to the complexity of the image deblurring technology,the existing methods mostly stop in the simulation stage of PC platform and adopt different methods with different degradation model.The deblurring result depend on the degradation model,long time-consuming,and the poor effect on image deblurring.Therefore,this paper takes general unified image deblurring methods as the research object,in the base ofstudy image deblurring algorithms principle,combined with high performance multi-core DSP.The characteristics of embedded software and hardware platform were researched and the image deblurring software system were designed and implemented.The multi-core DSP processor TMS320C6657 selected in this paper adopts TI Keystone architecture,it is high performance and low power to be used as an arithmetic core to construct image deblur system hardware platform with high performance characteristics.This paper analyzes the system function,including image data reception,image deblur processing,processing result output,image output display and other functions.The hardware platform of data path design,Ethernet interface design and extended memory interface design was studied.Combined with multi-core DSP hardware and software technology and universal unified image deblur algorithm,This paper implements an image deblur system that based on Heterogeneous Multi-core Communication and Shared Memory Model.This system combines the advantages and characteristics of multi-core DSP,and makes use of optimization methods such as C code optimization,compiler optimization,SIMD inline function optimization,Cache-EDMA multi-buffer queue optimization.After optimization,the computing time of the multi-core DSP platform is greatly reduced,and the multi-core operating efficiency is improved,which makes the system not only from the traditional PC,but also the system miniaturization and wider application range.