Research On Multi-exposure Images Fusion Algorithm Based On Dynamic Scene

High dynamic range image processing capture effectively more detail information in a scene with a higher dynamic range than the camera capable of capturing with a single exposure and is able to more truly record natural scenes,and has a wide application prospect.Exposure fusion is a simple and efficient technology for generating a high dynamic range image by directly fusing the multi-exposure images sequence.When using multiple images,local motion of objects will cause the ghosting artifacts to appear in the synthesized HDR image,which seriously influences the quality of the fused image and reduces the visual effect of the image.To solve this problem,this paper makes in-depth reasearch and analysis,then designs and realizes the ghost-free multi-exposuse images fusion based on dynamic scenes in two different ways.This paper first introduces the key technolygy of high dynamic range imaging and analyzes the multi-exposure images fusion based on static and dynamic scenes in detail.The quality evaluation of image fusion result is summarized.Aiming at the problem of ghosting artifacts in dynamic scene,a multi-exposure image fusion algorithm based on bitmap motion detection is studied.and an improved ghost detection and elimination algorithm with no reference image is proposed.The motion bitmap is generated by the median threshold bitmap and photometric relation.The double detection technology improves the accuracy of the detection of ghost area.The zero mean normalized cross correlation is used to guide the correction of weighted map,which extends the dynamic range of the moving region.Finally,laplacian pyramid blending is done to get the final enhanced image.Aiming at the problem that the dynamic range of the fused image is not completely extended,a new multi-exposure fusion algorithm based on moving object registration is proposed.Firstly,the least square fitting is applied to get the relation of pixel-wise intensity matching,the intermediate images is genereted by the bidirectional intensity mapping function,the moving object is detected on the super pixel block and replaced by the intermediate images,sequence replicas of original image are obtained,consistent with the reference image structure and the input image color.Then,the traditional multi-exposure images fusion method performed in the RGB color space cannot manipulate the intensity information,and a static scene multi-exposure image fusion method based on IHS space is proposed,which transforms the input low dynamic range images from RGB color space to IHS color space and deals with the intensity and chrominance information separately.The fused intensity component is achieved by weighted sum in multi-resolution fashion and the weight map is estimated by the quality measures: local contrast and saturation.For the saturation and hue components,we choose the best pixel on the same location using the ‘winner-take-all' manner to recover the fused chrominance.Finally,the multi-exposure fusion algorithm of IHS space is used to fuse the image replicas after registration.the proposed method preserves more detail and richer color information while eliminating the ghost.To verify the effectiveness of algorithm,several multi-expouse images with different motion scenes are blended by the proposed methods and compared with several excellent deghosting fusion algorithms.The evaluation and analysis of results is completed in subjective visual effect and objective quality of information entropy,average gradient and spatial frequency.The experiment results demonstrate the two proposed deghosting methods have advantages and disadvatages.The proposed method can effectively eliminate the ghosting artifacts in the dynamic scenes and produce the enhanced image with rich details and good visual effect and have a certain practical application value to the high dynamic range imaging.