Improved Image Enhancementalgorithm And Realization On Arm Platform

Image is an important source for people to acqurie information than otherforms of information for its content is more vivid and specific.However, in theactual case image acquired is inevitably subject to a variety of factors, makingthe image quality reduced, which not only affects visual effects in somedegree, but also brings some difficulties in subsequent processing. For imageacquired in fog weather and insufficient light conditions, we study two kindsof enhancement algorithm for degenerated image. Main work is as follow:1.For the fog weather image, the limitation of the dark channel priorlaw is that it is invalid in the sky area and similar areas, resulting in colordistortion in the image sky area, and the transmission refinement process isrelatively time-consuming. We make dynamic compensation for thetransmittance in the sky area, and also present an approximate method toaccelerate the transmission refinement process. The experiment results showthat these improvements can effectively eliminate the color distortion in thesky area, and greatly accelerate overall program speed on the condition ofsatisfying basic visual effect.2.For low-light image, the disadvantage of virtual exposure model is that the fixed standard deviation can not meet the requirements of image noisein different conditions, and low-light image also have low signal-noise ratioespecially in dark areas, tone-mapping function enhances the imformationdetails while makes noise amplification stronger, resulting in poor credibilityof the image information in these areas. We introduce a noise estimationmethod which can estimate image noise adaptively in different conditions,meeting noise reduction requirements under different noise levels, and we alsouse a piecewise tone mapping function to solve the over-enhanced problem,making overall effect enhanced while noise amplification suppressed in darkareas.3.Finally we build an image accquistion and processing system basedon ARM platform, and then realize the two improved algorithm on the system.The experiment results show that algorithm can be used in practicalapplication.