Illumination Recovery And Decomposition For Single Image

With the development of multimedia technology,image becomes one of the most common visual information.The image is generated by the interaction of object material and illumination.Illumination is one of the most important information in the image,and it is very important for the production of images in the scene.The colour of the object in the image depends on the illumination in the scene,and the colour of the same object is not the same under different illumination conditions.However,due to the influence of illumination and the surrounding environment,the effect of illumination in the image does not necessarily satisfy people's needs.Therefore,the follow:ing issue need to be resolved:how to make effective illumination editing and processing to create new images according to users,needs.Illumination recovery and decomposition for image is an image processing technology which is put forward under this circumstance.As the key technologies of image illumination editing,illumination recovery and decomposition for image has attracted widely public attention.It becomes a hot spot of research nowadays.The shadows in image can be used to estimate object geometry information and lighting information in the scene.Therefore,this paper carries out a thorough analysis on the point of shadow processing.It discusses the theory,method,key technology and application of the illumination recovery and decomposition for im-age.Specifically,the research is carried out in three aspects:shadow removing for image with complex textures,shadow removal for soft shadow image and illumination decomposition for image with multiple light sources.This paper has the following three main contributions:(1)We propose a novel shadow removal algorithm based on illumination optimization,which is used to recover the illumination in shadow regions.Based on the dis-tribution of the illumination in image,the paper builds an illumination recover framework to recovering the illumination in shadow regions.Due to the complexity of the image texture,the illumination recover is processed on a local patch.Then the brightness is optimized in all shadow regions by using the relationship between local patches to get the high-quality shadow removing result,which has a consis-tent illumination in the image.In addition,it can recover the detail information of texture in the shadow area effectively during the process of the shadow removing.This algorithm can process image with complex textures.(2)We develop a new soft shadow removing method based on the multi-scale decom-position of the image.By using the characteristics that the illumination in soft shadow regions is variation,the algorithm proposes a smoothing method based on the related pixels to complete the multi-scale image decomposition.Then a local to global illumination optimization strategy is constructed,which is used to re-move the shadows in the image.In addition,the shadow boundary distortion parts are recovered by using image texture syntheses methods,leading to a natural and authentic shadow removal result.(3)We propose an illumination decomposition method for image with multiple light sources.The method presents a novel coarse-to-fine strategy to decompose the input image into multiple components corresponding to different light source.Firstly,the paper reconstructs the lighting environment of the image,such as the numbers and position of the light sources in the scene.Then,using the illumination cues from shadows,we estimate the coarse illumination,decomposed image emitted by each light.At last,the coarse decomposed results are fined by using the presented light-aware illumination optimization model which produces better and more convincing illumination decomposition result.The texture details under shadows are also be recovered.Focusing on editing and processing illumination of image,the paper studies the image illumination editing and decomposition from different perspectives,of which the main line is the shadow in the image,and provides detailed solutions to the key technical difficulties in each problem.The experimental results show that the proposed algorithms are effective and can be extended to a variety of image-related applications.