A novel fusion-based method for single backlit image enhancement
AbstractIn this work, a new simple but effective fusion-based strategy for enhancing single backlit image is proposed. The fundamental idea of this paper is to blend different features into a single one to improve the specific quality of image. First, three derived inputs are generated by luminance improvement and contrast enhancement on the original value layer of the HSV space. Second, one weight is designed to emphasize feature in the derived inputs. Finally, an adjustable value layer is obtained by fusing derived inputs and weights in a multi-scale procedure. The new HSV space is transformed into RGB space to produce the final enhanced image. Since different types of information are effectively blended, the enhanced image achieves a decent trade-off between luminance improvement and contrast enhancement. In addition, the proposed fusion-based method requires only one input image and has a relative low execution time since most operations are pixel-wise, suitable for practical applications.
From left to right: original image, MSRCR, CD method , SECEDCT method  and the proposed method.
Reference C. Lee, C. Lee, and Chang-Su Kim. "Contrast Enhancement Based on Layered Difference Representation of 2D Histograms," Image Processing, IEEE Transactions on 22.12 (2013): 5372-5384.
Top: original images. Bottom: our results.