Multi-Scale Exposure Fusion Based on Multi-Visual Feature Measurement and Detail Enhancement Representation

Abstract

Multi-exposure image fusion (MEF) is used to generate a high-quality image from a series of images with different exposure levels. The multi-scale-based MEF method achieves better fusion performance than single-scale-based method because it can better preserve the global contrast information. However, this method still has the problem of the loss of details in the fused results. To solve this problem, a novel multi-scale exposure image fusion method based on multi-visual feature measurement and detail enhancement representation in the IHS color space is proposed. First, three visual features of the source multi-exposure images, namely contrast, saturation, and exposure, are measured and are then adopted to construct the initial weight maps with adaptive weighting coefficients. Second, to optimize the initial weight maps and obtain the middle weight maps, a decision map construction method is proposed by comparing the pixel values of the detail maps of the intensity components, which can enhance the representation of detail information. Third, guided filtering is applied to eliminate the noise in the middle weight maps to obtain the final weight maps, which improves the visual effect of the fused image. Finally, image pyramid decomposition and reconstruction are performed on the source images and the final weight maps to achieve the final fused image. Numerical experimental results indicate that the proposed method outperforms state-of-the-art methods in terms of subjective visual and quantitative evaluations.