Fuminobu Komura

Geometric correction algorithms for satellite imagery using a bi-directional scanning sensor

High-precision geometric correction algorithms for a bidirectionally scanned image that are based on separate geometric models according to the distortion characteristics are described. The modeling accuracy is evaluated through the use of simulated and observed Landsat Thematic Mapper (TM) images. TM images include a discontinuous distortion and a high-frequency distortion due to attitude fluctuations and bidirectional scans. A geometric model consisting of a single-valued-function is used to correct these distortions. The iteration calculation of the mapping cannot be converged, resulting in an enormous amount of calculation and data. Therefore, geometric models consisting of separate geometric models of the bidirectional scans and frequency components are proposed. >

Accelerated skew coordinate transformation for high resolution SAR image formation

An outline of accelerated skew coordinate system is explained for producing high quality images from the synthetic aperture radar (SAR) signals and an algorithm to perform coordinate transformation for this coordinate system is proposed. In SAR image formations, the SAR signal has been transformed into the skew coordinate system in order to correct spatial distortion, known as range walk, which causes the degradation of the range resolution of the SAR images. However, this transformation destroys the space invariance of the signal parameters in the azimuth direction. The accelerated skew coordinate system can correct the range walk without destroying the space invariance. The effectiveness of the proposed coordinate transformation algorithm for the SAR image formation is verified by the experiments using simulated SAR data and actual SEASAT SAR data.

A fast and accurate algorithm for SAR Doppler center frequency estimation

A fast and accurate algorithm to estimate Doppler Center Frequency (DCF) automatically from spaceborne Synthetic Aperture Radar (SAR) raw data is presented. Firstly, the principle of DCF estimation is examined and it is pointed out that the speckle nature of the SAR image plays an essential role. Secondly, the origin of estimation error by the conventional method which utilizes azimuth power spectra of raw SAR data is analyzed on the basis of the principle stated above. Thirdly, a new algorithm which is fast and accurate is derived from the estimation error analysis. Lastly, the comparison between conventional methods and the proposed algorithm is made using SEASAT-I SAR data, and the effectiveness of the algorithm is confirmed.

Interpolation algorithm for unequally spaced image data with image quality correction capability

An image interpolation algorithm considering the sensor aperture characteristics of unequally spaced image data is presented. Recently, it has become possible to gather images by a high spatial resolution sensor on a satellite. Precision image processing is required to correct images which have blurs and geometric distortions due to the sensor characteristics. To realize precision image processing, we propose an unequally spaced interpolation filter with image quality correction capability. It consists in a restoration filter, and equally and unequally spaced interpolation filters. Image quality improves by about 15 percent in unequally spaced regions, and interpolation accuracy of less than 0.5 in gray levels is obtained. This is shown by evaluation experiments using thematic mapper sensor-type simulation data.