Image Enhancement Algorithms Based On Guided Filter

Image enhancement is a technology to preserve the detail and improve the visual effect of the image.With the development of digital imaging equipment and processing technology of image,there is a higher requirement on image quality.Therefore,image enhancement algorithm has become an important research direction in the field of image processing.This paper focus on the tone mapping algorithm and multi exposure fusion.The ordinary image has only 0-255 gray scale for each pixel.When representing a natural scene with a large range of brightness,it will lose a lot of detail information.High dynamic images can solve this problem.Although high dynamic images can store brightness values accurately,they cannot be displayed on standard display devices.Tone mapping is a technique that compresses high dynamic range(HDR)images into low dynamic range(LDR)images so as to display them on display devices.In this paper,an adaptive tone mapping algorithm based on guided image filter is proposed.We present a simple adaptive compression operator to compress the input HDR image according to its average luminance.Images with higher average luminance are more compressed while images with lower average luminance are less compressed.What's more,in the same image,the brighter pixels are compressed more than the darker pixels.With both the average luminance and the contrast taken into account,the compression operator we proposed is able to adapt to different HDR images,which is much more accurate and perfect.After preprocessing the input HDR image by the adaptive compression operator,a LDR image is obtained.In order to preserve more detail information,the LDR image is decomposed into a base layer and a detail layer to process separately.Compared with other tone mapping algorithms,not only can our method preserve more detail and reduce halo artifacts,but also it is adaptive and requires no parameter setting.To compare the performance of tone mapping algorithms objectively,the image quality assessment method FSITM and TMQI are introduced.The experimental results show that the proposed algorithm leads to better output images compared with other six tone mapping algorithms.Multi-exposure fusion is a method to obtain an image with a proper exposure from a set of images with different exposure of the same scene.The high-exposure images preserve the detail of the dark part in a scene,while the low-exposure images preserve the detail of the light part in a scene.The purpose of the multi-exposure fusion is to fuse the detail of each images into one image as much as possible.We make a study on three traditional multi-exposure fusion algorithms and make an improvement of the image fusion algorithm based on guided filter.The original method only takes the saliency of images into account when computing the initial weight,which is crude and inaccurate.However,the improved method also takes the colorfulness into consideration.The more colorful the image is,the bigger its weight is.From the subjective comparison of the experimental results and the image quality assessment proposed by Hossny,it is clear that the improved method is better than the original one.