Remote Sensing
Remote Sensing
PAN-sharpening fuses the fine spatial detail of a high-resolution panchromatic image with the rich spectral information of a lower-resolution multispectral image.
The goal is to produce high-resolution color imagery that preserves both sharp structures and accurate spectral characteristics
Relative radiometric and spatial calibration improve the consistency and sharpness of satellite imagery. Relative radiometric calibration corrects detector-to-detector response variations through non-uniformity correction (NUC) and denoising, while spatial calibration optimizes focus and evaluates image sharpness using modulation transfer function (MTF) measurements. Signal-to-noise ratio (SNR) and related image-quality metrics are also measured to verify stable sensor performance.Â
Image denoising removes sensor- and system-induced artifacts such as random noise, vertical stripes, and periodic wave patterns from satellite imagery. Effective methods suppress unwanted signals while preserving edges, textures, and spectral information needed for reliable analysis.
SAR-to-EO translation converts difficult-to-interpret radar imagery into visually intuitive EO-like color representations. By learning structural and contextual relationships across modalities, it improves interpretability while preserving important scene content despite speckle noise, geometric distortion, and local misalignment.
Image mosaicking aligns and blends multiple overlapping satellite images into a single seamless, wide-area product. It corrects geometric differences, selects low-visibility seamlines, and balances radiometry so that roads, coastlines, and other structures remain continuous across image boundaries.
Super resolution reconstructs a high-resolution image from one or more low-resolution observations. It estimates missing high-frequency details while preserving natural edges and textures, enabling clearer visualization and more accurate downstream computer-vision analysis.
Computer Vision
Frame rate up conversion increases a video's temporal resolution by generating intermediate frames between consecutive inputs. Accurate motion estimation and motion-compensated interpolation reduce judder and blur, producing smoother playback while preserving object boundaries and fine details.