Multiple Exposure Image Fusion Considering Camera Motion And Moving Object

With the rapid development of image processing technology,the research on high dynamic range image is becoming richer and richer.Additionally,the application field has extended to many fields,such as games,photography,image compression and so on.However,the current research problems are too narrow,and the standardized experimental data are very scarce and fragmented,which is not conducive to the research of the systematic HDR image algorithm.In this paper,the three most common problems,camera motion,moving object and fusion quality are studied in the HDR production process,and more complex data are collected to complete the experiment task.The paper mainly includes three aspects:(1)Using geometric registration to solve camera motion.Utilizing image mapping function,SURF operator and RANSAC to get the image affine transformation matrix so as to realize the preliminary geometric registration experiment.The geometric displacement of the data taken by the camera fixed by tripod is small,so it can be completely corrected.However,the data taken by the handheld camera still have irregular distortion after initial registration.(2)Moving objects will result in ghosts on the HDR image.In this paper,the region of the moving object is corrected by the combination of IMF and PatchMatch.Then the motion state of the moving object on the low dynamic range image is consistent.At the same time,it can also correct irregular distortion.(3)Determining the saturation,contrast,and exposure quality as the image quality evaluation factor.Transform from the traditional pixel-level weighted fusion to the region-based fusion.Using the super pixel segmentation to generate the homogeneity region and calculate the evaluation factor of regions.Inspired by the definition of HOG feature,contrast is replaced by the texture complexity of the area.Adding the edge enhancement factor,the fusion results are obtained by the pyramid multi-level image fusion.This paper uses two kinds of data,respectively taken by camera fixed by tripod and handheld camera.The experiment shows a complete set of HDR production process.The experimental result conforms to the requirements,and it has a certain degree of improvement compared to the prior improvement.Meanwhile,the problems existing in the algorithm are also analyzed.