Image-to-image registration is applied to register overlapped images of a scene from different dates and views. Area features of the same regions are extracted from images as matching entities.The boundaries are matched in the frequency domain (Fourier descriptors) as area features. Two applications to extract the features from images as matching entities, "Edge enhancement and thinning" and "Region growing" are examined for this method. Registering a pair ofremote sensing data using this method is presented to show the secular change and the seasonal change of the data.

View full abstract

In the previous studies aerosol's optical thickness and Ångström exponent were retrieved from the satellite measured reflectance data by assuming pre-fixed complex refractive index of aerosol. In this study, for the first time, we estimated aerosol's real part of complex refractive index, together with its optical thickness and Ångström exponent from the directional reflectance and polarization data measured by POLDER on board ADEOS satellite by three different retrieval methods, namely, RP-Method, RR-Method, and PP-Method. RP-Method uses both directional reflectance and linear polarization values in a single infrared band. On the other hand, RR-and PP-Methods use directional reflectance and linear polarization values in two infrared bands, respectively. We found that RP-Method gives the best results in comparison with sky observation data at our study site, synchronized with the overflight of ADEOS. Finally, we presented the retrieved distribution maps of aerosol's three optical parameters over the Japan Sea and the western part of the Pacific Ocean by RP-Method.

View full abstract

In this work we describe procedures for detection of clouds and water/ice discrimination of the cloud top using the combined information of the polarization data from ADEOS/POLDER (0.865μm) and the infrared data from OCTS (10.9μm). On the basis of the satellite images obtained on November 10, 1996, we show that the cloud detection algorithm for POLDER proposed by CNES tends to overestimate the cloud regions. Better estimation can be obtained by adding a brightness temperature test by means of the OCTS data. A single-scattering simulation, carried out for water droplets and hexagonal ice crystals, indicates that the phase discrimination of the cloud top can be accomplished by comparing the polarized reflection data of POLDER at two scattering angles of 100^°and 140^°. The result is in reasonable agreement with the classification with the OCTS temperature data.

View full abstract

An algorithm is presented for the atmospheric correction of satellite data in the visible and near-infrared spectral region. The algorithm consists of analytic equations which are described only with the radiance components easily obtained with a radiation transfer code such as MODTRAN3. A three-step approach is introduced to precisely evaluate the pixel reflectance. In the first step, the path radiance is removed and in the second and the third steps, the adjacency effect from the neighboring pixels is removed. Using this method, a sensitivity analysis is performed for the NOAA AVHRR channel 1 and channel 2. To obtain a test image, we consider the multiply scattered component and the indirectly ground-reflected component, both of which are dependent on an average reflectance around the target pixel. It is found that a precision of about 1 % (channel 1) and 0.5 % (channel 2) can be obtained with respect to the pixel reflectances if the same atmospheric conditions are applied both to the generation of the test image and to the retrieval of the reflectance image. The method is actually applied to AVHRR images with the optical thickness obtained from ground measurements synchronized with the satellite overpass. The present algorithm is suitable for cases in which the surface reflectance changes on a pixel by pixel basis. Since the algorithm is described with radiance components, its application to any type of satellite sensor is quite straightforward as compared with the conventional lookup table approach.

View full abstract