A Method of Edge Detection with LANDSAT TM Data

Abstract

Satellite remote sensing can obtain a wide picture of the earth surface at the same time. But, the data collected from the satellite position are affected by the atmospheric conditions due to the weather. In order to improve quality of the remote sensing image the authors introduce a sophisticated image processing technique, i.e., a connection method. We propose an edge detection and enhancement algorithm to decrease an error probability of the classification of earth surface objects as much as possible.
The remote sensing image reflects complex geographical structure. Hence, edges obtained from the corresponding LANDSAT TM data may become complex according to the fluctuation of the earth surface. Furthermore, edges included in the image have many branch points to separate the edges to different directions. In this case, it is difficult to apply directly a traditional method to detect the edges.
The connection method stated above is to detect edges on the image by applying the differential operator to the remote sensing data and thresholding the resultant output. Next, the detected edges are connected by the nearest neighborhood method. Applying the above method to the LANDSAT TM data on October 8, 1984, we show the effectiveness of our method to find the edges on the satellite remote sensing image.