Enhancing Sentinel-2 Images using Super Resolution
Enhancing Sentinel-2 Images using Super Resolution
Requirements
- M.Sc. in Machine Learning, Data Science, Computer Science, Mathematics, Telecommunications, or similar
- Knowledge of Python
- Software development skills
- Basic concepts of image processing
- Basic concepts of data science, concerning data analysis, processing and machine learning
Description
This thesis focuses on developing an efficient and accurate deep learning-based Super-Resolution (SR)technique specifically tailored for enhancing Sentinel-2 satellite imagery. By leveraging time series of optical, and possibly SAR data, the goal is to enhance the openly available Sentinel-2 imagery by increasing its resolution. This capability is crucial for various applications, including land cover classification, urban classification, and natural resource management. The proposed methodology will involve extensive exploration of available datasets, such as SAtlas and Proba-V, and state-of-the-art superresolution methods like SRCNN, ESRGAN, and SR3. Through comprehensive analysis and experimentation, a novel deep learning-based algorithm will be developed and implemented to address the challenges of superresolution in varying geographical areas. The thesis will encompass the following activities: (i) thorough investigation of datasets and existing methods, (ii) design and implementation of a deep learning model tailored for superresolution, (iii) rigorous evaluation and comparison of the proposed technique with established approaches, and (iv) visualization of results to demonstrate the effectiveness and applicability of the developed methodology.