The "information explosion" has brought about wide dissemination and usage of digital image and video. Content-based research focuses on ways to retrieve image of interest by the visual content of image. This thesis presents an outline-based method for image content analysis and representation. First, algorithms are proposed for extracting high information outline segments from image. The information content of the outline segments is analyzed using entropy and self-information of the image. Experimental results show that the proposed outline segment extraction algorithm captures the visual content well. Then, a partial signature method is proposed for image representation. The partial signature method reflects partial information content of outline and it allows variation in different parts of the images. Algorithms are proposed for similarity comparison of images using the partial signature.