Abstract
The objective of this study is to design the Feature Enhancement method of the satellite multispectral data applying Genetic algorithms (termed the FEG method), and to investigate its applicability. The feature images, such as contrast, edge information in the satellite image are useful to interpret the land cover condition, lineament feature, fault and fracture pattern, etc. However, due to the reduction of the amount of information after generating the image features, there are some limitations to enhance the feature image itself for displaying, and the quality of enhanced image is not guaranteed in some cases. So, in making various kinds of feature images, the amount of information of the image should be increased in some way. In this study, the enhancement method, which increases the "Entropy" of the feature image (as a fitness function in GA), was designed by applying GA operations. Through the several experiments, we conclude:
1) The GA operation functioned well to increase the entropy of the feature image itself, based on the transformation process of the FEG method.
2) It was found that the amount of information, as well as the quality of the feature images, could be improved by using the FEG method. The image generated with the FEG method should contribute to the image interpretation by the specialist working on the various kinds of research fields using satellite multispectral data.
3) Furthermore, it was confirmed that the FEG method works well not only for a contrast image, but also other feature images such as the standard deviation image and the prewitt filtered image. The expandability of the FEG method is expected for the enhancement of other image features.
