Blue Sky Detection for Content-based Television Picture Quality Enhancement

Using high-resolution flat panel displays for TV systems, the visibility of signal impairments has grown. As a solution, content-based image processing can be used for obtaining higher levels of picture-quality improvement, thereby outperforming traditional TV image-enhancement methods. This paper presents a new algorithm and feature model for blue sky detection, which enables content-adaptive enhancement of TV video sequences. The algorithm analyzes the image, creates adaptive position and color models, and classifies the sky areas of the image using a pixel-accurate soft segmentation. Such a probabilistic measure matches well with the requirements of typical video enhancement functions in TVs. We have evaluated the proposal for typical and critical natural scenes, and obtained a clear improvement over state-of-the-art techniques.