Chinese Journal of Computers   Full Text
Title	Feature Preserving Coupled Bidirectional Flow for Edge Sharpening and Image Enhancement
Authors	FU Shu-Jun1),2) RUAN Qiu-Qi2) MU Cheng-Po3) WANG Wen-Qia1)
Address	1)(School of Mathematics and System Science, Shandong University, Jinan 250100) 2)(Institute of Information Science, Beijing Jiaotong University, Beijing 100044) 3)(School of Aerospace Science and Technology, Beijing Institute of Technology, Beijing 100081)
Year	2008
Issue	No.3(529—535)
Abstract &   Background	Abstract In the past decade there has been a growing amount of research concerning partial differential equations in image enhancement. In this paper, a feature preserving coupled bidirectional flow is presented, where an inverse diffusion with a soft edge decision is performed to sharpen edges along the normal directions to the isophote lines (edges), while a normal diffusion is done to remove noise and artifacts(jaggies) along the tangent directions on the contrary. The two converse diffusion forces are split into a coupled form to stop the cancellation between each other. To preserve image features, the nonlinear diffusion coefficients are adjusted according to the local differential geometry of image. Experimental results demonstrate that the algorithm substantially improves the subjective quality of the enhanced images. keywords image enhancement; edge sharpening; bidirectional diffusion; anisotropic diffusion; shock filter; inverse flow; differential geometry background Image enhancement and sharpening are important operations in image processing and computer vision. Main information of an image resides in such features as its edges, local details and textures. Image features are not only very important to the visual quality of the image, but also are significant to image post processing tasks, for example, image segmentation, image recognition and image comprehension, etc. Therefore, it is crucial to preserve and even enhance image features when one removes image noise and sharpens edges at the same time. Taking above into account, the authors think that image enhancement is composed of two steps: Features detection and the processing by corresponding tactic according to different features. In the past decades there has been a growing amount of research concerning partial differential equations (PDEs) in image enhancement. The PDEs based image processing techniques use synthetically such modern mathematical tools as partial differential equations, differential geometry, vector analysis and field theory, computational fluid dynamics, bounded variation space and the theory of viscosity solutions. Its basic idea is to evolve an image, a curve or a surface in a PDE model, and then to obtain the desired results by solving the PDE numerically. Virtually, after having been effectively discretized numerically, the partial differential equation is just transformed into a nonlinear local iterative filter. In this paper, incorporating anisotropic diffusion with shock filter, the authors present a geometry-driven nonlinear coupled bidirectional flow equation to remove image noise, and to sharpen edges by reducing their width simultaneously. This work was supported by the Natural Science Fund of Shandong Province, P.R. China (No.Y2006G08); the Researcher Fund for the Special Project of Beijing Jiaotong University, P.R. China (No.48109); the Open Project of the National Laboratory of Pattern Recognition at the Institute of Automation of the Chinese Academy of Sciences, P.R. China; the General Program Project of School of Mathematics and System Sciences of Shandong University, P.R. China (No.306002). The authors have finished a lot of researches on nonlinear bidirectional flow equation in image processing. This work is another important effort among the projects. It will produce great influence in aspect of theory and arithmetic in the field.