Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 102 === Based on multiple low dynamic range (LDR) images with different exposures, multiexposure image fusion computes weighting maps of the LDR images to generate a high dynamic range like (HDR-like) image. In this study, a multiexposure image fusion approach is proposed. Within the proposed approach, contrast limited adaptive histogram equalization (CLAHE) and homomorphic filtering are used to correct the intensity of each LDR image, cross-image median filtering is used to generate reference image, and gamma correction is used to compensate saturation loss. Then, weighted least square (WLS) optimization and the L0 smoothing filter are used to extract details by parameter which estimated by just noticeable distortion (JND) with reference image. Weighting maps for each LDR image are computed by visibility and cross-image consistency and refined by cross bilateral filter. Finally, an HDR-like image is achieved by multiresolution spline based scheme. Base on the experimental results obtained in this study, the performance of the proposed approach is better than those of four comparison approaches.
|