所報 104

13/38

－　　－11Published PaperMulti-Frame Image Processing of Himawari-8/AHI DataToshiro SUGIMURA, Yuuki UCHIDA, Sadayoshi AOYAMA, Tomohito ASAKA and Keishi IWASHITAHimawari-8 is Japan’s next-generation geostationary meteorological satellite using a highly improved sensor, Advanced Himawari Imager (AHI). The satellite observes all of the Earth’s surface (within visual range) every 10 min and all of Japan every 2.5 min. Therefore, 24 images of Japan can be observed sequentially in an hour.Tasks performed by Himawari-8, operated by Japan Meteorological Agency (JMA), include weather monitoring by collecting information about cloud coverage and movement. Clouds appear, move, and dissipate slowly, and this process can be observed using Himawari-8/AHI sequential images. On the other hand, because many objects on the ground are fixed, Hiamawari-8 also provides multi-frame images of immovable objects on the ground.The authors apply high-resolution image reconstruction using multi-frame low resolution images in this paper. This technique was developed to enhance video images. Since a set of Himawari-8/AHI sequential data is considered to be a multi-frame image, high-resolution reconstruction processing was examined as an enhancement opinion.Comparing 24 images observed sequentially in an hour does not yield noticeable distortion among them. If the pixels in each image are divided into 10×10 subpixels, several distortions, particularly in Band 3, become apparent because the resolution in Band 3 is higher than that of other bands. If the distortion of images divided using subpixels corrected automatically, subpixels can be fixed in their positions relative to each other to produce an enhanced image.This study examined the feasibility of multi-frame image processing using Himawari-8/AHI data observed sequential-ly over an hour. By adjusting the distortion of multi-frame images by subpixel comparison and automatic registration, it is possible to improve image quality from multi-frame image processing using mean value, median value, or a regres-sion function to shift subpixel positions in an image. As a result, Himawari-8 data can be improved to observe more in-formation than provided by the raw data.Keywords: automatic registration, multi-frame image, super resolution technique, Himawari-8ひまわり８号連続観測データを使った画質の向上について　2014年10月7日、次世代静止気象衛星「ひまわり8号」が打ち上げられた。観測波長帯数、空間分解能、観測時間でそれぞれ機能が向上している。特に日本域等特定の領域は2.5分間隔で連続観測され、種々現象の動きが観測できる。地表面は雲と違い移動することが無いため、これら連続観測された画像は対象を複数のフレームで撮影していると言い換えることが出来る。このことから、画質の改善が可能と思われる。また、各画像間に若干の位置ずれがあるとすれば、解像度の改善も可能であると考えられる。　本研究では、連続観測された画像間の位置ずれをサブピクセルに分割して調査し、それぞれの画像間に最大0.5画素のずれが存在し得ることを確認した。位置ずれを自動探索し、高解像度化および高画質化について検討、調査したところ画質の向上について良好な結果が得られた。キーワード：位置ずれ、マルチフレーム、高解像度化、ひまわり8号Journal（掲載誌）Transactions of the Japan Society for Aeronautical and Space Sciences Vol.16, No.3, 224-229, May 2018