Design And Implementation Of Embedded Video Defog System Based On Retinex Theory

The acquired image is severely degraded caused by fog and haze,which greatly reduces the usefulness of video surveillance equipment.Now video defogging system generally use optical lens or a hardware device for defogging,but its high price causes the video defogging system to become more expensive.Therefore,in considering reducing the cost,design a low-cost video defogging system it is meaningful.Based on the ARM platform,using ordinary CMOS camera and software MSR algorithm ways to replace the existing high-cost hardware defogging solution which can reduce video defogging system cost.Thesis mainly works is as follows:1.Do a detailed analysis of existing technical solutions at home and abroad video defogging system,simulations and compare the existing advantages and disadvantages of commonly used image enhancement algorithms in defogging applications,considering several algorithms and in effect defogging difficult to achieve on ARM embedded platform,choose a defogging video system solution based on Retinex theory.2.Under the solution,the completion of the hardware platform.To reduce system cost,thesis in the hardware design,does not use hardware defogging device,implemented in software defogging algorithm;camera with CMOS cameras instead of the optical lens fog lenses,in order to achieve when mass production cost control,the application of market maturity CMOS sensor chips and lenses that he design a CMOS camera;processor selects Renesas RZ/A1 H processor,which built large capacity 10 M RAM,the system can reduce the need for external memory;in other peripheral chips to meet demand,Select the most cost-effective components.3.On the hardware platform,system software design.It consists of four parts: an image capture;image-based Retinex theory defogging;JPEG image compression;Ethernet transmission.During image processing,since the image pixels and more,a large amount of system operation,in order to improve overall system speed defogging,the entire code optimization.Take the MSR lookup table for the logarithm of the number of hours used;Gaussian convolution of the original image and the template will be converted to floating-point arithmetic multiplication shift calculation,shaping calculated results are also easy to find a table of logarithms;in the image reflection when taking the index components,linear stretch approach to take in place of index calculation;color when color space conversion,will be converted to floating-point multiply-shift operation and avoid floating-point operations,and the use of ARM NEON image acceleration technology,the first performance a plurality of pixels color space conversion while;in JPEG compression stage using FDCT transformation,accelerate the conversion speed DCT.4.To complete the system test on the hardware platform,including camera image quality tests,images in the JPEG compression test,test to monitor indoor and outdoor defogging effect test,the test results meet the design requirements.Through testing,the defogging system designed by a software-based video technology,is able to adapt to different fog environments.It have a better effect on capturing the target of the low-speed movement,and is able to adapt to the needs of large video surveillance occasions.