How to reuse existing annotated image quality datasets to enlarge available training data with new distortion types.
Tomas MizdosMarcus BarkowskyMiroslav UhrinaPeter PoctaPublished in: Multim. Tools Appl. (2021)
Keyphrases
- image quality
- training data
- image distortion
- training dataset
- quality metrics
- peak signal to noise ratio
- image quality measures
- human visual system
- visual quality
- data sets
- distorted images
- perceptual quality
- perceptual image quality
- test data
- test set
- standard test images
- quality assessment
- reconstructed image
- image fidelity
- image quality assessment
- bit rate
- supervised learning
- decision trees
- image resolution
- high image quality
- compression ratio
- image data
- training set
- pixel values
- benchmark datasets
- learning objects
- image quality metrics
- motion estimation
- active learning
- learning algorithm
- machine learning