Research On Threshold Segmentation Of Non-uniform Illumination Image

The extraction and segmentation of the target image under non-uniform illumination is one of the most important and difficult points in the image processing.But in the actual process of image taking,due to uneven illumination conditions,images taken out will produce such as high light region and shadow region and irregular edge region,which affects the quality of information and read the picture.Aiming at the above problems,this paper proposes a segmentation method based on improved adaptive threshold,by adding noise to estimate the background image,and separated by an improved filtering algorithm for image filtering and correction to achieve the target and background.The experimental results show that the method can quickly and effectively segment the foreground and background,and extract the effective target information.The main research work of this paper is as follows:(1)An improved filtering algorithm is proposed.An improved mixed median filtering algorithm is proposed to deal with the problem of the median filtering algorithm in dealing with the probability of noise.The first element in the filtering window in different directions to take three times median,then three and a median median of a total of four times the median calculation,finally obtained is mixed median.The effectiveness and superiority of the improved filtering algorithm are verified by comparing with the original median filtering algorithm.(2)A new adaptive threshold segmentation method is proposed.Due to the uneven illumination,the global threshold algorithm will be affected,and the local threshold algorithm has some problems such as slow real-time and poor robustness.In this paper,an adaptive threshold method is proposed,which is divided into blocks according to the position of pixels,and then a threshold T is selected for each sub image.In order to determine the advantages and the effect of the algorithm,we choose the contrast image to establish the corresponding image library,at the same time,we compare the above algorithms.The test is not difficult to find that the adaptive threshold segmentation algorithm to edge features of the objects in a more complex environment,can make the non-uniformillumination image effectively,depending on the volatility of the reasonable threshold to confirm whether the fluctuation,can meet the needs of various forms of segmentation.(3)A threshold segmentation system is designed for a non-uniform illumination image.In this paper,the basic principles of the system design,the main steps,the organizational structure,the layout of the various control of GUI and the implementation of each module are described in this paper.Using a graphical user interface with MATLAB to complete a non-uniform illumination image segmentation system,the threshold segmentation algorithm and other main experiments together,you can complete the image into the image preprocessing and image segmentation function,form a simple pre read image,interactive operation way of image processing image processing and image segmentation and visual system.