Abstract: An image restoration method based on noise gray scale difference estimation is proposed. In order to ensure the quality of restored punctate image, the image is convoluted respectively by various directional uniform arithmetic operators, and the results from convolution are subtracted respectively from the original images. The estimation image of noise gray scale difference is constructed by extracting the pixels with minima of gray scale differences. Then, the intensity and the position of the spot in the image can be got, and the spots are smoothed by neighboring pixels. Test results show that noise gray scale difference estimation can express exactly intensity and position of the spots in the punctate image with small operation amount needed. The proposed method has been used experimentally to restore three dimension biological microscopic image, and its results show that the lighteness of spots is controlled efficiently and small spots are wiped off in restored image. The image restoration with super-resolution is achieved with good effect.