Research Of Shadow Detection And Removal Algorithm Based On Single Image

When obtaining a required image, it is always in a bad situation that the quality of image will decline because of the influence of shadow. It is a phenomenon that Shadow is the cause of degradation caused by the imaging conditions, which results in the information lost or disturbance in the target image, weakening the image of optical information, reducing the interpretation accuracy and speed of the image. Obviously it will seriously affect a variety of quantitative analysis and the application of the image.Shadow is generated due to the block of a light source by the object, means the loss of image information will have an effect on the subsequent image matching, feature extraction, pattern recognition, etc. so the research on shadow in a image and the shadow process are very meaningful. In this paper, shadow detection and removal algorithm aimed at single image had been investigated, and the focus was put on the feature-based shadow removal approach. Firstly, it summarized the background and significance of algorithms of the shadow removal, comparing the current research of shadow detection and removal algorithms in and abroad, and then the introductions of image forming theory, types and characters of shadow. The theoretical basis of shadow detection algorithm is the nature of the shadow, this paper draws on the ratio map method, in which the luminance information is particularly important in the shadow detection theory. Considering that the shadow area is usually the connected domain with larger area, and the brightness value of the shadow area is generally smaller compared with the well lit area, this paper proposes a shadow detection algorithm based on the average brightness and region growing method. Firstly, the need for such a shadow image in RGB space is to make early preprocess to obtain contrast-enhanced gray-scale image, and then gray image is binarized, based on the binary image a coarse center point of the shadow area is located, which can help us find the average value of brightness in HIS space. Finally, with the help of the average value of brightness and the shadow center obtained in the previous step, it is able to do the region growing and edge swell, resulting in a final detection of shadow area. This algorithm made use of the brightness information and the color information of the shadow center point of the targeting image, which obtained a good detection result matching well with the real shadow area, improving the automatic detection accuracy of the algorithms. In the end, The effectiveness of the detection algorithm had been verified by experimental simulation.Shadow detection is the precondition of shadow removal, and the effect of shadow removal directly depends on the detection result. On the basis of shadow detection, a shadow removal algorithm based on sub region matching is proposed, which can realize adaptive illumination transfer between shadow and non shadow regions. This algorithm can deal with different texture and brightness conditions of the shadow image, so that the shadow area of the pixels in the color, brightness, texture to the lit area to restore the effect of light and the more continuous surrounding scene. experimental result demonstrated the capabilities of our algorithm in both the shadow removal quality and running performance. Finally, in the C++/GUI platform, the program can realized the shadow detection and removal process through the design of the software system, completed the whole implementation process of shadow removal.