Research On Shadow Information Enhancement Algorithm Based On Multi-scale Retinex From Remote Sensing Images

In recent years,with the launch of artificial satellites,remote sensing technology has gradually been applied to many fields,such as meteorological observation,geographic mapping,urban planning,resource and environment monitoring,etc.Through remote sensing technology,people can interpret remote sensing images to obtain the information in the images and make use of it.However,due to the influence of geographical environment,the sun’s rays are blocked during the propagation process to form shadows by mountains,rivers,trees or tall buildings in the city.The existence of shadows has advantages and disadvantages.On the one hand,we can use shadows to estimate the height or shape of buildings,but on the other hand,the presence of shadows will cause the loss of information in the shadow area of the remote sensing image,such as texture information,color characteristics etc,reduce the quality of the image and affect subsequent image detection,target recognition,classification and other operations on remote sensing images.Therefore,in recent years,in order to solve this adverse impact,many experts and scholars at home and abroad have proposed research topics on the shadow area processing of remote sensing images.The processing of the shadow area of the remote sensing image is mainly divided into shadow detection and shadow information enhancement.Shadow detection can extract the contour information of the shadow area,which is a prerequisite for the shadow information enhancement operation.And shadow information enhancement is to improve the image quality by restoring or compensating the information in the shadow area to facilitate the recognition and extraction of remote sensing images.However,due to the small shadow area and wide distribution in some remote sensing images,it is easy to produce errors in the shadow detection and enhancement of remote sensing images,so the main research purpose of this dissertation is to reduce the color information and detailed information loss,enhances the contrast and details information of images as much as possible.This dissertation mainly selects the karst landform from Guilin,Guangxi as the research object,and this dissertation mainly explores the shadow information enhancement algorithm of remote sensing images in this area.The landform of Guilin is a world-famous karst landform,and its surface features are mainly peak forest-depressions.Due to this landform,the shadows in remote sensing images in this area are mostly mountain shadows.The characteristics of such shadows are many and small,and the distribution of mountain shadow relatively evacuated and will confused with river features easily,it has a certain impact on the operation of shadow detection,shadow enhancement,feature recognition,information extraction,or feature classification.Therefore,this dissertation uses this as an object of exploration to enhance image shadows information,which forms a strong contrast with non-shaded areas,paves the way for subsequent image processing.In this dissertation,through comparative study of existing algorithms,this dissertation proposes shadow information enhancement algorithms suitable for the research object of this dissertation.(1)By comparing the advantages and disadvantages of the shadow information enhancement algorithm based on HSI color space and the multi-scale retinex shadow information enhancement algorithm,this dissertation proposes a multi-scale retinex shadow information enhancement algorithm combined with HSI color space.Most of the existing multi-scale retinex algorithms process RGB three channels independently.Although the color distortion problem in the single-scale retinex algorithm has been solved,but the relationship between the RGB three channels has been destroyed,resulting in overexposure or underexposure,and halation.Combined with the multi-scale retinex algorithm of the HSI color space,the components of the RGB color space are first mapped to the HSI space,in the HSI color space,the brightness component is enhanced,and the saturation component is adaptively adjusted according to the correlation coefficient,and finally processed image back to RGB color space to obtain the enhanced image with shadow information.Using this algorithm not only solves the effect of the water in the research object in single-scale retinex algorithm,but also solves the color distortion problem in the multi-scale retinex algorithm,and achieves a better shadow information enhancement effect.(2)In the HSI color space,brightness and hue cannot be completely separated.Therefore,for this problem,this dissertation proposes a Lab-based multi-scale retinex shadow information enhancement algorithm.Based on the method in Chapter three,the image is transformed from the RGB color space to the Lab color space.In the Lab color space,only the L component is in multi-scale retinex processing,which can not only change the brightness component,but also ensure that the chroma channel value remains unchanged,without causing excessive chroma distortion.