Font Size: a A A

Research And Design Of Colorization Algorithm For Night Vision Image

Posted on:2022-01-17Degree:MasterType:Thesis
Country:ChinaCandidate:S WangFull Text:PDF
GTID:2518306731977809Subject:Computer technology
Abstract/Summary:PDF Full Text Request
The rapid development of science and technology has driven the widespread application and spread of images.Compared with gray-scale images,people are more willing to view color images.Clearly,colorful high-contrast image can help people understand the scene and identify targets faster.However,it is difficult to make a low-illuminance image consistent with the daytime color.In the current night vision technology,the grayscale image is obtained at first,and night vision image colorization technology is then applied.Image colorization technology is widely used in military,film and television,medical,national defense,and surveillance fields.Obtaining a natural night vision image has important research significance and application value.This article focuses on the colorization technology of night vision images including infrared images and visible light images as the main research content.The specific work content can be summarized as follows:1)We propose a colorization method for night vision images based on guided filter image fusion.This method uses the edge preservation and image smoothing characteristics of guided filtering to optimize the weight map to obtain a fused image with complementary details.The proposed method first decomposes the input single-band grayscale image into two scales,then constructs a weight map based on the source image,and then weights and fuses the weight map to generate a fusion image in two scales.Finally,the reference image is used to realize the color transfer algorithm to perform the color migration of the fused image,and obtain a natural color night vision image,which better realizes the color fusion of the infrared image and the visible light image.Experiments show that,compared with the traditional image colorization algorithm,the colorized image obtained by the proposed method in this paper has prominent thermal targets and higher image contrast.2)We propose a colorization method for night vision images based on edge preservation.Aiming at the problems of single-band night vision images such as poor detail performance,low contrast,and blurred edge contours,the proposed method starts from two perspectives: image noise processing and edge extraction.On the one hand,the median filter is used to eliminate noise,on the other hand,in order to highlight the details of the image edge contour,the Laplacian operator is used to extract the detailed features.Then the two types of enhanced images are colorized using color transfer algorithm,and then color fusion is performed to obtain the final natural color image.The experimental results show that the color images acquired by this method are less disturbed by noise.The contour details of the target object are prominent,and the visual effect is good.
Keywords/Search Tags:Guided filtering, Night vision image colorization, Laplacian operator, Color transfer
PDF Full Text Request
Related items