Multi-Image Denoising And Its Application

With the continuous advancement of image-related technologies,digital image and digital video technology has been widely used in all aspects of human production and life.In the process of collecting digital image signals,once the ambient background light is insufficient,such as shooting at night and shooting under high magnification,the captured image has low brightness and low contrast,and usually contains a lot of noise.Noise seriously destroys image information,affecting image target recognition,and seriously affecting people's access to complete information.Therefore,research on how to remove digital image noise is still of great significance.At present,many imaging devices are very convenient to obtain multi-frame images.In order to obtain better image denoising effect,this paper studies related to multi-image denoising.The thesis includes the following parts.Firstly,the background and application value of image denoising are introduced.The related image denoising methods at home and abroad are reviewed.The current image denoising algorithm is analyzed and the shortcomings of single image denoising method are studied.Secondly,a multi-image non-local mean image denoising method is proposed based on non-local mean.First,the pre-denoising image and the method noise are obtained by using multiple noise image linear weighting combined with multi-image nonlocal mean filtering.Then,Gaussian filtered method noise and pre-de-noise image are used to solve multi-image non-local mean weights,and the final denoised image is obtained by combining weights and input images.Thirdly,based on the observation that increasing the number of image frames is beneficial to further remove image noise,this paper proposes a multi-scale multi-image fusion denoising method based on grid flow.Performing a downsampling operation on the input noise image sequence to obtain different scale images,and determining the reference image and the non-reference image according to the image sharpness.All the largest scale images are meshed,the image features are extracted,and the feature points are weighted with the mesh vertices to obtain a precisely aligned image sequence.Then,the reference image and the sequence median image are used to obtain a similar pixel set,and the set is scale-transformed to obtain a corresponding set of each scale.These sets are used to estimate the denoised images in the corresponding scales,and finally the fusion denoising is performed by using the mapping relationship between adjacent scale images.The innovation of this paper is to design two kinds of multi-image denoising methods,including(1)combining a multi-image and non-local mean filtering to design a multi-image denoising method,and studying the noise distribution of the simulated image and the natural image,and The noise effects are compared and the effectiveness of the method for denoising is verified.(2)A multi-image denoising method is designed by using grid constraints,grid flows and multi-scale.Grid constraints achieve accurate image alignment,grid flow achieves scale image alignment,brightness constraints achieve multiple image fusion at the same scale,and texture constraints implement image fusion at different scales.The experimental results show that the proposed method has good denoising effect and the operation speed is better than the traditional denoising algorithm.