BANding Detection

Banding, a visual artifact appearing as staircase-like contours in images and videos, is a significant quality issue stemming from compression and quantization processes. Current research focuses on developing accurate no-reference banding detection models, often leveraging frequency domain analysis and employing convolutional neural networks to identify and quantify banding severity, sometimes incorporating large, newly-created datasets of banded images with associated subjective quality scores. These advancements improve objective image quality assessment and enhance the user experience by enabling better detection and mitigation of this common visual distortion.