Research On Super Resolution Reconstruction Of Infrared Images

In nineteenth Century,F.W.Herschel discovered infrared radiation,which is a kind of electromagnetic radiation with wavelength between 0.75um?1000um.Infrared imaging refers to the infrared radiation energy which is invisible to the naked eye is captured by other infrared detectors and then converted into visible images.Infrared imaging technology can be classified into active and passive from the perspective of imaging methods.Considering that active infrared imaging needs infrared radiation source to assist in actual operation,which increases the difficulty of shooting,passive infrared imaging technology is mostly used in current research.Passive infrared imaging uses the infrared detector to directly receive the difference between the radiant heat of the target and the background to form an infrared image,so passive infrared imaging is also called thermal imaging.Passive infrared imaging technology is widely used in military,security and other fields because of its good concealment,strong penetration and high identification.Limited by the infrared band and camera equipment,the resolution of infrared image is low.In this paper,the infrared image super-resolution reconstruction algorithm is studied from three aspects of image interpolation,image enhancement and depth learning.Due to the lack of detailed information of infrared image,the super-resolution reconstruction technology using interpolation method is prone to produce sawtooth and ringing phenomenon.In this paper,the bilinear interpolation algorithm is improved based on the mean filtering principle,and a super-resolution reconstruction algorithm combining bilinear interpolation and local mean is proposed,which can further improve the super-resolution reconstruction performance of bilinear interpolation algorithm.Image super-resolution reconstruction is divided into two aspects: image size enlargement and image enhancement.The primary purpose of image super-resolution reconstruction is to improve image resolution and ensure image quality.In this paper,an infrared image enhancement algorithm based on region adaptive multi-scale strong light fusion is proposed,which has good effect in infrared image enhancement.The current image super-resolution reconstruction algorithm based on deep learning is limited to the structure design of convolution layer.This paper studies the channel image of each convolution layer,and proposes an improved convolution neural network infrared image superresolution algorithm,which can improve the reconstruction performance of convolution network.