Research On High Dynamic Range Image Signal Restoration Aenhancement Algorithm

Capturing images with cameras can record visual information of scenes,which is the most basic and important link of visual information analysis.However,comparing with the real scene irradiance,the dynamic range of ordinary cameras is far less.When the dynamic range of the scene to be captured is inconsistent with camera has,it will lead to the areas in the captured image appear overexposure or underexposure,and further result in detail loss in highlight or shade areas.While,the high dynamic range(High Dynamic Range,HDR)image can display the bright and dark scenes information on a common device as much as possible,and conforms to the perception of human eye.With the rapid development of high-definition digital industry,HDR images obtain a wide range of applications in the industrial image processing,machine vision,3D digital entertainment and other fields,becoming one of the hotpots in the digital domain.This thesis goes into study the HDR image and multi-exposure fusion technology on theory method,technical characteristics and the research actuality.Firstly,to obtain more scene information through single HDR images reconstruction,a single high dynamic image reconstruction method based on fuzzy theory is proposed.Secondly,taking into account that limited scene information can be obtained from a single image,then multiple exposure image fusion methods are concerned,thus,the thesis presents multi-exposure HDR image reconstruction algorithms based on multi-scale detail fusion and human visual adaptation functions,so that the details in bright and dark regions are kept as many as possible,and minimal visual distortion is achieved.Meanwhile,in order to achieve the effect of real-time applications,a gray-level exposure fusion algorithm is proposed,and the solution to resolve the ghost problem during the fusion process is given as well.Finally,a comprehensive evaluationmethod is proposed based on fuzzy theory for the reconstructed results.The main work and innovation of this thesis included:(1)Single high dynamic range image reconstruction.Since the dynamic range of camera sensor is usually inconsistent with the actual dynamic range of light,the main target dynamic range distribution is concentrated and forms a ultra-low dynamic range,thereby further leading to signal suppression.To solve this problem,the low dynamic range image is transformed from spatial domain into fuzzy domain through membership function,and the contrast can be improved by the enhancement fuzzy operator.Finally,the processed image is inverse transformed into the spatial domain,and then high dynamic range image can be achieved,which can restore the scene in the dark details,and retain the contrast of the image.(2)Multi-exposure HDR images reconstruction base on multi-scale detail fusion.This thesis maps the decomposed weight Gaussian pyramid through Dirichlet function,and assigns maximum weights for areas with rich information,then the HDR image can be reconstructed by the Laplace pyramid,which contains the maximal detail information and minimal distortion.(3)Multi-exposure HDR image reconstruction based on human visual modeling.Unlike other multi-exposure fusion algorithms,this thesis extracts the brightness information of each pixel directly from the HDR image sequence captured in the same scene with different exposures,and then the mathematical model of brightness sequence curve is established through the visual adapted S-shaped curve,on the basis of which,the discriminate method for best imaging pixel is given,so that the complex calculations of HDR generation and tone mapping are avoided,and the HDR image can be synthesized rapidly for display on a conventional device.(4)Multi-exposure HDR image reconstruction based on gray scale mapping function modeling.For LDR image sequence of arbitrary size,visual adaptation S-shaped curves need to be fitted that is the same as the number of gray-scale,rather than the camera resolution pixels.The HDR can be achieved by fusing the best imaging values directly,which greatly improves the efficiency of the algorithm fusion and can achieve real-time requirements for dynamic scene.The ideal state of multi-exposure image can be achieved by the design of the gray level mapping function;the ghost can be eliminated through moving target detection with difference method;finally a high dynamic range image can be achieved which can reflect the real scene information and not affected by the ghost.(5)Evaluation method for multi-exposure HDR image reconstruction based on the theoryof fuzzy.The proposed method takes into account single factor indexes such as image information entropy,average gradient,moderate exposure,mutual information,structural similarity index metric,cross entropy and so on.Based on the fuzzy comprehensive evaluation method,we get a comprehensive assessment index,which can reflect the small change of single indicator and meanwhile overcome its one-sidedness.