What is 'Super Resolution' in computer vision, and how is it achieved?
Instruction: Define super resolution and describe techniques to improve image resolution in computer vision tasks.
Updated
Example Answer
Super-resolution estimates a higher-resolution image from lower-resolution input. A pixel-loss model favors reconstruction fidelity, while perceptual or adversarial objectives can favor visually convincing texture. That texture is inferred, not proof that the original scene contained it. I would choose the objective from the use case and evaluate both fidelity and perceptual quality. In evidence-sensitive imaging, a sharper-looking result must not be treated as a recovered measurement.
Make it your own
Illustrative evaluation: compare a simple interpolation baseline and the learned model on held-out degradations; retain originals and inspect fine text, edges and unusual textures for invented detail.
Why this works
Perceptual quality and distortion can trade off. A realistic-looking pattern may be wrong. Follow-up: which downstream decision would become unsafe if an inferred detail were accepted as fact?
Sources: The Perception-Distortion Tradeoff.