Abstract: Most existing methods employ simple algorithms to process images,and they are effective to suppress image noise,however,the useful edge and detailed signal information of the image are also destroyed.To resolve this problem,the effect of different noises on gridding is analyzed firstly,and the grid line number and area root-mean-square(RMS) error are put forward as indices to realize the quantitive evaluation.Then the mathematical morphology algorithm is combined with maximum-between-clusters-variance method to suppress the highly fluorescent noise.The proposed algorithm is applied to denoising DNA microarray images with birth defect and compared with other conventional methods.Experiment results confirm that the presented method is practicable and efficient,and has good ability to reserve useful information.