Image thresholding method based on uneven lighting

According to the poor effect of segmentation problem of traditional segmentation method in the disposing of the text image because of the uneven illumination, this paper proposes a new method of image segmentation. First, the method calculates average value of original text image, with the average divided by image matrix of each element, it will get equilibrium matrix. Then, according the pixel numbers of text gestures in the text image to determine the Laplace template, combining this template and the original image convolution operation to get filtering image, and using the filtering image matrix to multiply the homologous elements in equilibrium matrix, it will get a relatively uniform background illumination image. Last, disposing the image with threshold segmentation, and eliminating hot pixel according variance, it will get a perfect segmentation result. The method has been simulated, and the results show that for non-uniform illumination text image, this method can obtain better segmentation effect.