Multiexposure Fusion Based on Multiscale Edge Information Measurement and Adaptive Pyramid Optimization

Multiexposure image fusion (MEF) technology aims to extract visible information in image sequences of the same scene with different exposure levels and combines the extracted information to generate a clear and visible composite image. In traditional methods, constructing weight maps and extracting fine details are two core issues. To overcome these problems, a novel MEF method based on multiscale edge information measurement (MSEIM) and adaptive pyramid optimization (APO) is proposed in this article. First, an MSEIM method is designed, which can extract the edge information using an improved multiscale morphological filtering technique, and the extracted multiscale information is merged by an adaptive weighted sum to obtain the edge maps. Second, weight maps are constructed via comparing the values of the pixels in the obtained edge maps. Third, an APO method is proposed to obtain an edge-preserving smoothing pyramid (ESP) from the weight maps via Gaussian pyramid construction (GPC) and adaptive guided filtering. Finally, the fused result is generated through multiresolution mixing on the Laplacian pyramid of source images and the ESP. Numerous experiments conducted on public datasets showed that the proposed method is superior to many state-of-the-art (SOTA) MEF methods in visual perception and quantitative analysis.