An automated method for identifying images in an image database based on scores assigned to a plurality of input keywords. Each input keyword is assigned a keyword score based on the number of images in the database that are associated with that keyword. Each image in the database is then assigned an image similarity score based on the keyword scores of the input keywords associated with that image. If a user selects an image in the image database and requests to see similar images, the keywords associated with the selected image are used as input keywords. Images in the database with image similarity scores indicating greatest similarity are provided to the user.