1995 年 15 巻 5 号 p. 419-431
The objective of this study is to determine two parameters (instrument function, damping factor) in the RGS-method (the image restoration method with Gauss-Seidel algorithm) as a image quality improvement method, which was published in the Journal of the Remote Sensing Society of Japan in 1992. The RGS-method can be easily carried out adjusting only two parameters for the satellite multispectral image of various quality. As a present procedure, two parameters are determined through the simulation for image quality improvement using RGS-method. But a satellite multispectral image shows various quality, it is important to know how to decide the value of parameters before the image improvement processing.
Recently, it is reported that the fluctuation index used in the fluctuation theory is useful for analysing correlation of visual complexity or human mind etc. First of all, the applicability of the fluctuation index for determining two parameters was investigated.
According to the examination for MESSR. TM and HRV data, in the case the image quality is good, a very interesting result was shown that the scatter of the points on the scatter diagram of the power spectrum gives essentially straight line. Further investigation was carried out to clear the causality between the image quality and linearity measure in the scatter diagram.
The procedures for determination of two parameters consist of four steps as follows:
STEP-1) Simulation of image quality improvement using RGS-method.
STEP-2) Determination of damping factor (Df) by means of fluctuation index.
STEP-3) Determination of damping factor (D1) by means of linearity measure.
STEP-4) Decision of damping factor by taking the average value between D, and Df. The results of this study are as follows:
1) The fluctuation index as the gradient for the power spectrum is calculated. Damping factor (Df) can be decided just in case the value of fluctuation index is changed remarkably, because at this time, it was confirmed that the spectral feature of improved image is not preserved.
2) As a results of the examination on the scatter of the power spectrum, the linearity measure shows a change of local spatial frequency feature. Damping factor (D1) could be decided by means of the maximum value for linearity measure.
3) When the image quality is good, it was confirmed that the value of damping factor is between D1 and Df Based on this result, damping factor is decided by taking the average value between D1 and Df. In conclusions, it was shown that parameters of RGS-method can be set automatically on the computer processing.