Research And Implementation On High Dynamic Range Image Tone Mapping

With the development of computer science and technology, People have higher requirement for image, High dynamic range(HDR) image can show more detail features,and the majority of information in natural scene can be recorded by HDR image; The dynamic range of the HDR image is extremely broad, which can be able to reach six orders of magnitude, while the ordinary display device owns two orders of magnitude dynamic range, and the display devices for high dynamic range image specially is expensive which is not desirable in real life. Therefore, How to make a HDR image displayed on a conventional low dynamic range(LDR) display device is one of the challenges in digital image processing research.One of the method for addressing above-mentioned problem is called HDR image tone mapping, The purpose of HDR image tone mapping operator is compressing the dynamic range of the HDR image and reserving the detail of the rich features at the same time. This paper describes the tone mapping research background and the basic theory of HDR images, meanwhile, introduces several classic tone mapping algorithms. In this paper, we analyze the tone mapping algorithms the predecessors had put forward, which was based on local edge-preserving filter in details, and some improvement is made on the proposed tone mapping algorithms. Firstly, The input image is scaled into one base layer and some detail layers by using Local edge-preserving filter. Then, since the predecessors' algorithm is poor in overall image contrast, aiming at this shortage, an adaptive logarithmic transformation is carried out in base layer in order to compress dynamic range, which can reserve the details in highlight and dark area at the same time.In addition, during the processing of enhancing the detail layers, in order to make better result, many tests are carried out with different parameters, which is time-consuming and inconvenient, aiming at this shortage, an adaptive sigmoid function is carried out in detaillayers which can enhance the detail meanwhile suppress the sharp edges with larger gradient. Then all the base layer and detail layers are summed up to give the result.The improved algorithm is a good choice for HDR tone mapping because halo artifacts associated with other traditional filters are avoided and the resulted image contains all the details in highlight and dark area, which not only helps humans recognize the real scene, but also overcomes the shortcomings of traditional display devices and is more adaptive to the future research in computer vision.