The Research Of Edge-preserving Algorithm Based On Guided Image Filtering

With the development of image processing, the image processing technology is increasingly applied to some professional applications, such as the Intelligent Transportation System, the city security system, large image data real-time processing and so on. The image that obtained by the image acquisition device directly tends to be influenced by various objective factors, and its quality is difficult to meet the requirements of most algorithms. Therefore, we often need to use some image preprocessing method to improve the original input image’s quality. The edge information is the most important part of the complete information of the original input image. In image preprocessing stage, how to increasing image sharpness and retaining image edge while removing noise is one of the most important research goals. Some traditional image preprocessing algorithms, just like the average filter, the median filter, are unable to retain the image edge details while removing noise and smoothing image details. However, we can’t have our cake and eat it too. The existing classical edge-preserving algorithms always have good effect but low operation speed, such as the anisotropic diffusion and the bilateral filter.Guided image filter is a kind of simple edge-preserving algorithm, it can smooth the image at the same time reserving the edge of the image. This algorithm works so efficiently that can reach the O(N) level of computation complexity. Meanwhile, it can avoid the gradient reversal artifacts around the edge of the image while the bilateral filter can not. So, the guided image filter provides a new solution for many image processing applications and can be used widely in the field of image processing. However, the algorithm can not avoid introducing halo around the edge of image just like most of local filer algorithm do.In order to eliminate the halo artifacts produced by the guided image filtering, this paper proposes a novel improved guided image filter that combines gradient information of guidance image to find out edges and direct the process of guided image filter. The main jobs and innovation of this paper are as follows:1. Summarize the current research status of edge-preserving algorithm, generalize the working principle, advantages and disadvantages of different algorithms. The bilateral filter, the guided image filter and the weighted guided image filtering algorithm using laplacian-of-gaussian edge detector are described emphatically, and combine experimental data to analyze them.2. This paper proposes an improved guided image filter based on gradient information. Our method use the gradient information of guidance image to find out edges and emphasize them, then computing the weight-value to avoid the halo introduced due to excessive smooth near edges.3. Summarize the best parameter ranges and robustness of the guided image filter, the weighted guided image filtering algorithm using laplacian-of-gaussian edge detector and our method by contrastive experiment and quantitative analysis. Our method has greater robustness and better performance than the other two algorithms in many computer vision applications, including image denoising, detail enhancement, multi-exposure fusion, etc.