Knowledge Based Approach for Automated Digital Image Processing

Automatic interpretation of digital images is a difficult task. In this paper, we have proposed knowledge based approach for Landsat image segmentation and interpreted without any prior image dependent information. It includes an integration of image processing techniques, knowledge from domain experts and ancillary information such as previous maps of the study area. We discuss three important issues in automatic digital image interpretation are image registration, road detection and knowledge based segmentation. In this study, major land cover types are organized in a hierarchical structure. A complete knowledge based segmentation technique may consist of two stages. The first stage uses the proposed method to segment a Landsat image by spectral knowledge rules. The second stage then collects more area dependent spatial rules and the prior map information to perform a complete segmentation. Nagao and Mastuyama have developed a knowledge-based system, which performs a structural analysis of complex aerial photographs using a technique called segmentation-byrecognition.. One advantage of this method is its flexibility in applying it to geo graphically different areas. This work develops techniques for automating the process of IRS LISS-III image interpretation. The experimental results show that the proposed method can be successful in segmenting complicated Landsat image. Through this study, we believe that the proposed hierarchical method is a promising approach for Landsat image segmentation. In order to make the proposed method more reliable and powerful, in future research topics, these are being incorporated.