2003 年 42 巻 5 号 p. 48-59
The Noise Reduction Filter (NRF) is improved from a Local Maximum Fitting filter (Sawada, et al., 2000) that has been developed to remove noise such as cloud and haze from time series NDVI images and extract vegetation information on cultivated fields. The time series data of SPOT/VEGETATION is processed by NRF. Then, the processed data are evaluated by two methods. One is comparison with an annual NDVI profile. Another one is statistic analysis based on the peak time and the peak value of the annual NDVI profile. Furthermore, an examination is conducted about the most suitable number of the time series data to enhance the processing accuracy.
As the result, information on SPOT/VEGETATION is kept, and a noise element is removed. NRF is effective in the extraction of the time series property of NDVI and planting information from the image. As for the number of time series data, difference of the peak time is not shown clearly although one-year data are enough numbers to get the good correlation of peak values and four-years data are needed to extract smooth curves of NDVI changes.