Separable Geodesic Filtering For Image Processing

Based on the analysis of image characteristics,image filtering is a kind of image processing methods to regain the original image by using characteristic information.Separable geodesic filtering method for fast image denoising has many advantages on image denoising implementation.Because the linear translation-invariant(LTI)filtering algorithm use the Euclidean distance for obtaining actual distance of adjacent pixels,which can effectively guarantee the fast calculation within the class.Although LTI filtering is very simple and is used extensively in early vision processing,it still has some disadvantages.LTI filtering smooth the noise,but blurs important structures along with noise,and outliers exert large influence on filtered outputs.So,separable geodesic filtering method is considered to be adopted.Based on the intrinsic geometry of an image surface-geodesic distance which is used to measure the similarity of two pixels,this method can filter out noise with preserved edges and fine-scaled details.Moreover,its separable implementation is simple and fast,and can process efficiently color images as well as gray images.According to simulation results of image denoising,the proposed algorithm shows its great efficiency.Comprehensively in the research of various kinds of algorithm on the basis of the geodesic distance and the denoising effect of noised image,this paper proposes an improved geodesic distance fast image filtering algorithm:In this paper,firstly we discuss the geodesic distance in one-dimensional space and propose a fast algorithm based on the additivity of one-dimensional distance.Then it will be generalized to two-dimensional images,and a fast separable geodesic filtering algorithm will be proposed.The filtering method is simple and has a linear time implementation.Considering the structure information between ranks of images at large,the improved geodesic distance fast image filtering algorithm makes the information of images more consummate.The experimental results show that competitive results can be achieved in image denoising and joint filtering,and the filtering method is simple and has a linear time implementation.The better denoising effect comes out.