Abstract: Traditional segmentation method can only partition an image in a single granularity space, with segmentation accuracy limited to the single granularity space. This paper proposes a multi-granularity synthesis segmentation method for remote sensing images based on the quotient space granular theory. A quotient space model of multi-granularity image segmentation is discussed. Image data field is used to express the spatial
correlation of pixels, and fractal dimension used to enhance the capability of discrimination between artificial surface features and the natural scenes. The watershed algorithm and iterative self-organizing data analysis technique (ISODATA) are applied to the gray image, data field image and fractal dimension image to produce multi-granularity segmentation results. This paper proposes a specific quotient space synthesis algorithm for
multi-granularity image segmentation. Experiments show that the proposed method can take full advantage of the segmentation result in every granular space. The multi-granularity synthesis segmentation is effective and can produce more accurate segmentation than that of a single granularity space.