Low Illumination Image Enhancement Based On Multi-frame Fusion

The image obtained under low illumination conditions often have problems such as high noise,low contrast,and color distortion.In order to improve the visual effect of low-illumination images,this thesis deeply studies the low-illumination image enhancement algorithm based on multi-frame fusion.With multi-exposure fusion technology as the core,the registration and fusion algorithms of multi-exposure images are studied respectively.A complete low-light image processing process is formed and implemented on the Android platform.The main work of the thesis is as follows:1.In order to solve the problem of multi-exposure image registration in low illumination scene,an image registration algorithm based on local template matching is proposed in this thesis.Firstly,the robust feature points are extracted from the selected reference images.Then,the corresponding positions of feature points are searched in the other images by using similarity criterion.In order to improve the range and efficiency of matching,the image is decomposed into Gauss pyramids,and the matching operation is performed at each level.Experiments show that the number of matching points obtained by this algorithm is large and the distribution is uniform.2.Aiming at the problem that existing multi-exposure fusion algorithms can not effectively restore image details and color information,this thesis proposes a multi-exposure fusion algorithm based on Retinex theory and verifies the application effect of this algorithm in low illumination enhancement.Firstly,the multi-exposure image sequence is divided into illumination image sequence and reflectivity image sequence by efficient illumination estimation method.Then,the two image sequences are fused differently to obtain the reconstructed illumination image and reflectivity image respectively.Finally,the final fusion image is obtained by multiplying the two images.3.The multi-exposure fusion algorithm proposed in this thesis can be applied to low illumination image enhancement,but there are still some problems to be solved.This thesis improves the algorithm from two aspects.On the one hand,because the image captured in low illumination scene is seriously affected by noise,this thesis combines image enhancement with image denoising,improves the processing of illumination image and reflectivity image,and enhances the anti-noise ability of the algorithm.On the other hand,in real scene,the acquisition of multi-exposure image will be affected by many factors,for example,camera jitter and moving objects exist in the scene.Therefore,this thseis proposes a multi-exposure fusion algorithm based on single frame image.4.In this thesis,the proposed algorithm is integrated to form two low-light image enhancement schemes based on multi-frame fusion.It has been successfully implemented on the Android platform and developed a camera application that can achieve low illumination image enhancement.The test results show that the application achieves the expected results and performs well.