Research Of Image Composition Method Based On Illumination Constancy

The key processes of digital image composition is sequenced technically by extracting the target foreground from an image,and then putting it into a new background.As far as possible after composition,the new image would achieve the same degree of fidelity as the image naturally acquired.In the issue of image composition just mentioned above,the mismatch of illumination attributes between foreground and background is a typical technical problem,includes illumination intense,color temperature and illumination direction.The research work in this paper is aimed at the digital image composition under the consistency of illumination intensity.Digital image composition technology mainly includes the extraction of the target foreground,and the composition of the target foreground and the new background.In this paper,Bayes Matting algorithm based on tri-map and the segmentation algorithm based on graph theory and strokes are both analyzed.Through making a compared experiment between Bayes Matting algorithm and a typical segmentation algorithm based on graph theory,it shows that the latter has a high extraction accuracy of Alpha image.In order to make the edge of alpha image get smoother and softer,this paper adopts the guided filtering to optimize it further.In the part of composition,the principle and pros and cons of the classic Poisson Cloning are analyzed in details,and a method of image synthesis based on brightness cloning is achieved by combining Poisson Cloning and the segmentation algorithm based on graph theory.The results of experiments show that this method has greatly improved the color distortion of Poisson Cloning,but when there is a large brightness contrast in composition boundary,the brightness of the background will penetrate to the target foreground more or less.Since there is a close relationship between the singular value decomposition(SVD)of the image and the image pixel value,based on the image composition method of ?-channel,a new digital image composition algorithm based on matrix SVD and a segmentation algorithm based on graph theory is proposed as well.The evaluation in this thesis includes qualitative and quantitative one,for the latter,the method based on illumination component extraction is put forward according to the aim of research,which is intended to evaluate the illumination consistency of target image via computing the illumination component's similarity between background image and target image.The results of experiments show that the target image used by the method of this paper can achieve the closest illumination component with that of source background image,that's to say the method in this paper can get the target image with better illumination homogeneity.