liuguoyou/exposure_fusion

Exposure fusion is a technique that creates a single image with optimal detail from a set of multi-exposed images. As developed by Tom Mertens et al., the proposed algorithm computes relevant quality measures; Contrast, Saturation, and Well-Exposedness. These measures are then combined to create a weight map used to blend each of the multi-exposed images to a single image with best exposure.

Exposure Fusion

Introduction

Exposure fusion is a technique that creates a single image with optimal detail from a set of multi-exposed images. As developed by Tom Mertens et al., the proposed algorithm computes relevant quality measures; Contrast, Saturation, and Well-Exposedness. These measures are then combined to create a weight map used to blend each of the multi-exposed images to a single image with best exposure.

Description

As developed by Tom Mertens [1], the exposure fusion algorithm computes relevant image quality measures; Contrast, Saturation, and Well-Exposedness. These measures are then combined to create a weight-map used to blend each of the multi-exposed images to a single image with best exposure. The need for exposure fusion rises in finding optimal exposure settings during photography. Since this requires mastery of the exposure triangle which is difficult, requires a trade-off, and outright impractical for some scenes, techniques such as high dynamic ranging (HDR) and exposure fusion have been developed to obtain a desirable final image. Unlike HDR however, exposure fusion doesn’t require computation of a camera response curve or tone-mapping

Image stack: here, the multi-exposed images are combined into a stack of images to be processed
Quality measures: here, the (a) contrast, (b) saturation, and (c) well-exposedness are computed
Scalar Weight Map: here, the quality measures for Image[i] are combined and normalized
Blending: the weight maps and images are blended to obtain a final image  
Final Image: this is the final image with best detail using Multiresolution blending (Multires.)  

The image below shows the image stack, decomposed quality measures, and final weight maps.

Results

The images below shows outputs for different exposure-bracketed images.

Another :
Another :

Another :

A comparison with other methods: