Model-assisted content adaptive detail enhancement and quadtree decomposition for image visibility enhancement

In this work, a simple and unique, yet effective framework for visibility enhancement, using image decomposition, adaptive boundary constraint and quadtree-based dehazing, detail enhancement and fusion is proposed. The input image firstly undergoes image decomposition, to be split into its its ambient illumination and reflex lightness components. For increasing brightness and contrast, contrast-based enhancement algorithm is applied to the reflex lightness component and the ambient illumination component is dehazed through the application of atmospheric scattering model. In order to estimate the airlight, instead of using the whole image, a simple and efficient method based on quadtree decomposition is used. The transmission map is computed through contextual regularization using adaptive boundary constraints. The dehazed ambient illumination component is passed through detail enhancement for enhancement of sharp edges. The resultant image and the enhanced reflex lightness component are then combined together through fusion to obtain the final, artifact free, enhanced image with preserved colors and details. The proposed methodology is evaluated using numerous images and compared with 8 different state-of-the-art techniques. Visual and quantitative comparison of the proposed methodology with existing state-of-the-art techniques demonstrates the effectiveness of the proposed technique.

Automated summary

This work proposes a simple and unique framework for visibility enhancement using image decomposition, adaptive boundary constraint, and quadtree-based dehazing, detail enhancement, and fusion. The proposed methodology is evaluated and compared with state-of-the-art techniques, demonstrating its effectiveness.