Research On Image Detail Enhancement Algorithm Of Infrared Imaging System

Due to the characteristics of high dynamic range,blurred target details,reduced overall contrast,and significant noise,the infrared images generated by infrared thermal radiation signals are unable to meet the needs of certain projects in the context of engineering applications.Therefore,it has become the development trend of infrared imaging technology to retain the rich detail information in infrared images and to solve the problem of displaying high dynamic range infrared images on traditional displays.The purpose of this paper is the implementation of an algorithm for IR image detail enhancement and dynamic range compression,which will preserve and enhance the detail information in the image,avoid introducing various noises,and facilitate the subsequent subjective visual observation or machine processing.First of all,this paper presents a theoretical analysis of the mainstream algorithms related to infrared images based on the current state of research at home and abroad,and an in-depth study of the principles and key technologies of the corresponding algorithms,and summarizes the corresponding advantages,disadvantages and technical limitations through experimental simulations of the processing effects of the algorithms,so as to establish a reference for the performance evaluation of the algorithms in this paper later.Secondly,to address the halo artifacts in the edge regions of the image in the current algorithm,an infrared image algorithm based on the single-scale Retinex theoretical model and improved weighted bootstrap filtering is used for processing.The algorithm uses hierarchical processing to decompose the original image into irradiation and reflection components,linear mapping and histogram equalization to enhance the contrast and edge information of the irradiation component,and introduces an edge weighting factor in the local variance-weighted bootstrap filter to enhance the local edge information of the reflection component,and then fuses the components to obtain the final enhanced infrared image.Experimental analysis of two methods,subjective visual evaluation and objective technical index quantification,is performed by selecting the infrared open dataset images.The objective evaluation indexes of information entropy and average gradient are significantly better than other algorithms with the average value of 6.7341 and 14.6655,so it is concluded that the method is superior and robust in image detail enhancement.For the problem that image noise in real scenes can lead to image quality degradation,an improved two-stage nonlocal mean filtering based algorithm is used for image denoising.The method uses the structural similarity parameter of the image to improve the Euclidean distance of the nonlocal mean filtering algorithm,and the two-stage nonlocal mean filtering method is used to improve the image blurring and distortion problems that exist after the traditional nonlocal mean filtering algorithm is filtered.According to the experimental analysis based on the evaluation of subjective vision and objective indexes,the algorithm is more consistent with the visual characteristics of human eyes in terms of subjective vision,and the average values of two objective indexes,peak signal-to-noise ratio PSNR and structural similarity SSIM,are31.2723 and 0.8299,which are better than other denoising algorithms in terms of noise suppression and image detail retention.Finally,according to the hardware design problem of infrared image enhancement algorithm in practical engineering applications,the hardware platform design is implemented by selecting the core processor platform using FPGA technology for hardware porting of the algorithm in this paper.It is experimentally verified that with an average resource utilization rate of 37% of FPGA,the system in this paper obtains almost the same image processing quality as the software platform in terms of image visual effect enhancement,while maintaining good portability and more potential upgrade space.