Research On TV_L1 Model Based Image Fusion

Image fusion is a research hotspot in the field of image processing,which is widely used in infrared visible light image processing,medical image processing and other fields.The fusion method can synthesize the useful information of the same scene obtained by each sensor into a single image,which can give more accurate,more comprehensive and more reliable image information than a single image.This technique makes full use of the redundant information and complementary information in the source image,which is more in line with the visual characteristics of human or machine,and is convenient for target detection,identification and tracking.At present,neither the image edge nor the texture can be expressed accurately,so it is difficult to capture the target from different sensors into a complete image.To overcome this drawback,the total variation TV and partial differential equation PDF based image processing method has been developed and applied to the field of image fusion.In this paper,the image fusion model based on variation is mainly considered,and two improved fusion algorithms based on gradient transformation and total variation are proposed.The main tasks are as follows:1.To improve the visual effect under different background brightness,the human visual system(HVS)is added to the canonical term model.According to Weber's law,the response of human visual system mainly depends on the local variation of background brightness,and its mathematical expression is that the human eye can perceive the local difference in the image.Under different background brightness,different gradients have different visual effects.Experimental results show that the improved gradient transformation model can enhance the edge details of the fused image.2.To solve the step effect problem in integral order variational fusion,a parameter adaptive fusion algorithm for coupled gradient TV_L1 model is proposed by combining fractional calculus theory with dual theory.The fusion problem is transformed into the TV_L1model optimization problem.The adaptive parameters are obtained by using the fixed point iteration method.In the model,the fidelity term can keep the structure edge feature of the target,the regular term can keep the texture details of the background gradient change in the image,and the adaptive parameters can effectively balance the fidelity term and the regular term smoothing.The experimental results show that the parametric adaptive variational algorithm can effectively suppress the step effect,preserve texture and structure information,and have high convergence.3.To improve the clarity of the target image,the morphological open operation method is used to deal with the weight of the target image,then the infrared weight map(IWM)can be obtained.Experimental results show that the proposed algorithm has better performance in visual evaluation and objective evaluation compared with the existing fusion algorithms because it can retain more details and edge information in the fused image.