Research On Visible Light And Infrared Video Fusion Technology

With the development of imaging technology, the image and video taken by single sensor has been unable to meet human’s needs, which leads multi-sensor video fusion technology attract more and more attention. Visible light video and infrared fusion technology is an important part of multi-sensor video fusion, which plays a very important role in the military, industrial, medical, monitoring and other fields. The images captured by the visible light sensor contain abundant information of the scene,but they are easy influenced by the weather and illumination, in addition, they can’t detect hidden target either. Images captured by infrared sensor are little affected by the weather, illumination conditions, they are mainly used for the detection and localization of heat targets, but their detail and scene information reflecting ability is poor.Therefore, it is significant to do some research in infrared and visible light video fusion.This thesis focuses on the multi-decomposition and fusion strategy of multi-scale fusion of infrared and visible light video, and proposed a new fusion algorithm.First of all, this thesis analyzes the characteristic of infrared and visible light images,the fusion process of infrared and visible light images is given, and introduces several image enhancement algorithms. Through the study of several image enhancement algorithms, this thesis proposes a new image enhancement algorithm which is suitable for infrared images. Compared with other image enhancement algorithms, the algorithm proposed by this thesis achieves better results.Then, this thesis emphatically discusses the fundamental framework of multi-scale image fusion and image multi-scale decomposition methods, and studies the Pyramid transform, Wavelet transform, Curvelet transform, Contourlet transform and its improvement transform, non-subsampled Contourlet transform, and analyzes the advantages and disadvantages of the above several decomposition method, which help us find out the suitable method of multi-scale decomposition. On this basis, this thesis deeply studies the fusion strategy of low frequency and high frequency coefficients.First, we analyses the low frequency coefficients fusion based on fuzzy logic, and then,we proposed a new fusion strategy of low frequency coefficients based on the Structure Similarity. In the fusion of high frequency coefficients, in order to describe the degreeof prominence in some position, this thesis proposed the local prominence degree index,which is used for the fusion of high frequency coefficient.Finally, this thesis evaluates the images and videos fusion quality processed by the algorithms discussed in this thesis through the subjective evaluation and objective evaluation aspects. The results show that, the algorithm proposed in this thesis has a better ability than other algorithms in extract target information from infrared images and the scene information in the visible light images.