Phong model based binarization and its application to billet images with non-uniform illumination

In image processing and analyzing, binarization method can be used to extract the feature or target from the image. OSTU is a classical biarization algorithm. But for the non-uniform illumination image, the processing result by the classical OSTU is not suitable. This paper presents an improvement OSTU scheme for a class of non-uniform illumination images. The character of non-uniform illlumination is modeled by a Phong model and then used to adjust the OSTU binarization threshold. The proposed scheme is applied to the defect detecting of billet images with non-uniform illumination. Details of the engineering implementation are introduced. The experiments on the actual billet images show the effectiveness of the improved binarization method.