Research And Implementation Of Automatic Segmentation Technology For Illumination Highlights In Complex Scene

With the advancement of science and technology and the improvement of people's living standards,the application of effect images in complex scene in the fields of home and architectural design is more and more extensive.When the effect image with the highlight area is enhanced visually and directly,it is easy to cause overexposure of the area.Therefore,it is necessary to divide the highlight area and then enhance it separately with others,but manually dividing is time consuming and laborious.In the thesis,the effect image illumination highlight segmentation system is researched and developed.The system can automatically segment the highlight areas in the effect images.The main research contents and results of the thesis include:(1)Based on SLIC(Simple Linear Iterative Clustering)algorithm,the thesis designs a super-pixel method to automatically segment the effect image illumination highlight areas,a partial clustering method to automatically determine the number of super-pixels,and realizes automatic segmentation of illumination highlights.(2)Based on fuzzy C-means clustering algorithm,the thesis designs a fuzzy clustering objective function including brightness,color and distance parameters is designed to improve the weight of the brightness value in the clustering,and the edge fitting degree is improved.The segmentation method generated by the clustering method combined with the super-pixel biased clustering method can improve the effect of the highlighted region of the image after the segmentation.(3)The thesis designs an evaluation criterion for the segmentation of effect image illumination highlight area.The experimental results of segmentation method in the thesis,SLIC method and other traditional super-pixel method for segmenting illumination highlights are compared.The segmentation method in the thesis is applied to image enhancement,and experiments verify the effectiveness of the segmentation method.(4)The thesis designs and develops a software for effect image illumination highlight segmentation,which realizes the automatic segmentation of the effect image illumination highlight area.The software combines the super-pixel block shape rule,the uniform size and the high degree of fit of the fuzzy method,and achieves the segmentation of effect image illumination highlight area.The experimental results show that the software can effectively segment the effect image illumination highlight areas in complex scene.