Font Size: a A A

The Study Of Foreground Extraction And Fusion Algorithms Under Complex Background

Posted on:2015-09-13Degree:MasterType:Thesis
Country:ChinaCandidate:Q J LiuFull Text:PDF
GTID:2298330428951923Subject:Parallel and distributed theory
Abstract/Summary:PDF Full Text Request
Foreground extraction and fusion has been widely used in film production andbody feeling games. Now common methods we use include blue screen key,background modeling method and the method based on Kinect and so on. But there isstill no reliable method of the foreground extraction under complex background, so itis a hot spot in the current study. This paper studies the foreground extraction undercomplicated background and subsequent image fusion, improving the existingmethods on the basis of predecessors’ and putting forward a new method forforeground extraction under complex background.This paper firstly realizes the foreground extraction method based on blue screenkey. However, blue screen key method needs a blue background wall. Thisrequirement limits the scope of its use. To solve this question, some scholars putforward to model the complex background. This paper analyzes the two kinds ofbackground modeling method: background modeling method based on colorinformation and background modeling method based on Gaussian mixture model.Background modeling method based on color information can distinguish the shadowof foreground well and background modeling method based on Gaussian mixturemodel is a foreground extraction method with best effect. When the backgroundvaries slowly, these methods have very good effect on foreground extraction. Butwhen the change of background is complex, the background modeling methods can’thave good effect.Microsoft uses Kinect based on speckle to extract the foreground. Kinect usesinfrared speckle to get depth information of the scene and then extracts the foregroundunder complex background. But when the edges of the foreground are complex, usingKinect to extract the foreground is not accurate. So this paper puts forward the method that uses flood water filling to optimize extracting foreground with referenceto the source images.This paper proposes a new method to extract foreground. By irradiatingnear-infrared light on the foreground, it can distinguish foreground with the complexbackground. So it can extract the foreground well.Based on the extracted foreground of this new method, this paper discusses a methodthat makes it have a seamless fusion with the new background: Poisson equationmethod. Experimental results show that this method can create a new integratedimage.
Keywords/Search Tags:foreground extraction, Image fusion, Additional informationMethod, Poisson fusion
PDF Full Text Request
Related items