Abstract: A method for improving the quality of soybean images by reducing the effect of noise using multi-scale morphology is presented in this paper. First, multi-scale structure element was used to make open-and-close reconstructing operation to source image to construct morphological open-and-close towers; then different operations of morphological open-and-close reconstruction images between adjacent scales were computed to construct different towers of bright features and dark features. According to bright features and dark features of different scales, smoothed image was reconstructed. The proposed method has been tested on a set of soybean images corrupted with different types of noise. The results were compared with some other standard noise removal algorithms based on some standard performance measures. The results show that the method proposed in this paper is superior to others.