本研究使用空間解析度為 30 m 的 Landsat 8 OLI 近紅外波段影像。在此解析度下，常見人造斜邊目標難以提供足夠長度與穩定幾何條件的有效邊緣。因此，本研究探討在缺乏理想校準目標時，利用自然海岸線之大尺度邊緣估算影像銳利度。所提出方法結合影像退化模型進行模糊核估算，並透過離群值排除與統計篩選降低自然地物雜訊影響。此方法可降低尋找合適目標的門檻，並相較其他自然邊緣方法提升影像銳利度評估的準確度。;In recent years， declining launch costs have driven rapid growth in the deployment of low-Earth-orbit satellites. Many remote sensing missions now require not only high-spatial-resolution land observation， but also wider spatial coverage， higher revisit frequency， and improved temporal resolution. For applications such as target detection， tracking， and analysis， on-orbit image quality assessment is a critical issue in satellite payload validation and remote sensing applications， as the spatial resolution and sharpness of optical remote sensing imagery directly affect target identification and analysis.
Among various image quality metrics， the modulation transfer function (MTF) is widely used to describe an imaging system’s ability to preserve contrast at different spatial frequencies， making it a key indicator of spatial image quality. Traditional MTF measurement methods typically impose strict requirements on both test targets and observation conditions. They often rely on artificial targets or dedicated calibration sites with well-defined geometry， stable radiometric properties， and high contrast. For example， the slanted-edge method， which is commonly used for on-orbit MTF measurement， reconstructs the edge spread function (ESF) through sub-pixel sampling and subsequently derives the line spread function (LSF) and MTF. Although this method generally provides relatively reliable measurements near the Nyquist frequency， it still requires sharp edges with high contrast and uniform backgrounds. As a result， specially constructed high-contrast and highly uniform slanted-edge targets are often used to reduce the influence of background noise. However， the applicability of this approach is limited by site conditions and the physical size of the artificial targets.
This study uses Landsat 8 OLI near-infrared imagery with a spatial resolution of 30 m. At this resolution， common artificial slanted-edge targets are difficult to use， as they cannot provide effective edges with sufficient length and stable geometric conditions. Therefore， this study investigates the use of large-scale natural coastal edges to estimate image sharpness when ideal calibration targets are unavailable. The proposed method incorporates an image degradation model for blur kernel estimation and applies outlier removal and statistical filtering to reduce the influence of noise from natural land-cover features. This approach lowers the difficulty of finding suitable targets and improves the accuracy of image sharpness assessment compared with other natural-edge-based methods.
