Map Generation from Aerial Imagery Using Data Compression

In this paper, we propose a new method to generate a digital map from the aerial photograph. For map generation, it is necessary to develop a technique to recognize various naturaland artificialobjects in the image by computer processing. In general, a photo image has spatial and spectral informations. It is desirable that these two informations are effectively utilized. A new method proposed here integrates and utilizes them without complicated processing. First, the photo image is represented by a Minimum Spanning Tree (MST) and encoded into one-dimensional text, considering both spatial information and spectroscopic informations. Next, this text is compressed by using a dictionary technique. As the compressibility of text depends on the dictionary, the input image can be described as a n-dimensional compressibility vector using n dictionaries. Using 139 aerial imagery, effectiveness of this new method is illustrated.