Abstract: The gridding is a basic step in the gene chip image processing. To suppress the influence of the highlight stains which could interfere with the gridding of the high pollution gene chip images, the highlight objects were divided into three categories: spots, stain blocks and stain points. Based on its characteristics, every category was processed respectively and a new gridding algorithm was proposed. Firstly, the highlight objects were located by utilizing a global thresholding segmentation. Secondly, the strain spots were smoothed via image corrosion technology. The strain blocks were removed using a dilation technology to reduce the trace of edges. Finally, the autocovariance was utilized for gridding the noise-free gene chip images. Experimental results show that the proposed algorithm can obtain an ideal gridding result for the high pollution gene chip images and the spot detection average accuracy can reach 94.73%.