The Processing Of Background Difference Images

The paper mainly focuses on the processing of background difference images. With the rapid development of computer science and technology, the applications of color image have immerged in almost everywhere in industries. Color image contains much more information than grey image, so many traditional applications based on grey image were replaced by those based on colorful images. Recently, applications in video retrieval, face recognition and object fast tracing became common. Although color image possesses richer information than grey image, new difficulties come at the same time. The important lies in that the processing is affected easily by illumination variation. The characteristics constructed by color become more sensitive to illumination and change greatly with illumination variation. Therefore the recognition and processing of color image tends to be unstable. This obstacle affected the popularity of those applications based on color. So, the study on the relationship between color and illumination to disclose their stable inner relations so as to develop a robust method to deal with illumination variation is the most hottest area in this field. Uneven illumination including in time distribution and spatial distribution affect not only the senior feature of object color but also the low level on object recognition. Object area extracting and separating is base on feature recognition and feature measurement. The core of this method is image segmentation. So, color image segmentation is also a preliminary inspection. Due to the illumination infection, the color distribution will be stretched. As the result, it easily appears that one kind of color area is divided into some different areas in the segmentation process, especially the highlight area and gray area. Sometimes it leads to over-segment. Owning to color constant capability, the human vision system can restrain itself from over-segmenting while observing. This characteristic makes us to consider a possible solution to deal with the over-segmentation. This paper discusses a new method on area combination based on color continuous distribution. This method need not depend on the chromatic aberration between the areas, but consider the macroscopic trend of color distribution of the image areas. So it brings a good effect on overcoming the over-segmentation caused by highlight illumination. The results were achieved by using MATLAB, and the Wiener filter was used to improve the precision of object detection.