A background separation method of nonuniform image segmentation
Junqiang Liu, Jianming Gao, Qingming Shen, Junwei Tian · 2009
It is difficult to separate object from background using conventional method when the processed image is nonuniform. A new method is proposed in the paper for nonuniform image segmentation: Firstly, grid sample method is performed on initial image to reduce data space and prepare for background estimation. Secondly, Gaussian low pass filter (GLPF) is used to reduce the intensity value of the high-frequency points that is caused by objects included in the image. Thirdly, facet model based interpolation algorithm is used to estimate the background image. Finally, object image is acquired according to the difference of initial image and background image. Experiments were performed and according to the results the validity and adaptability of our method is enhanced obviously compared with conventional image segmentation algorithms.