Traditional algorithms based on graph theory were sensitive to the brightness of image, the time complexity was high, and in order to improve its segmentation effect, a new threshold segmentation algorithm, based on characteristics of human brightness perception, was proposed.This algorithm used the Normalized Cut criterion, and combined with graph theory and LIP (Logarithmic Image Processing) model. Meanwhile the algorithm utilized both intensity and position of information, had holistic character and its time complexity was acceptable. The experiments illustrated that the algorithm could more effectively distinguish an object from the background than traditional techniques, even in the condition of small changes in scene illumination.
