Quality-Oriented Perceptual HEVC Based on the Spatiotemporal Saliency Detection Model

Area of Science:

• Computer Vision
• Video Compression
• Signal Processing

Background:

• Traditional video coding like H.265/HEVC faces limitations in balancing bitrate and visual quality.
• Perceptual video coding (PVC) offers a promising approach to improve compression efficiency by considering human visual perception.
• Developing effective saliency models is crucial for enhancing PVC performance.

Purpose of the Study:

• To propose a novel H.265/HEVC-compliant PVC framework.
• To leverage a video saliency model for optimized video compression.
• To achieve significant bitrate reduction while maintaining high visual quality for high-definition video.

Main Methods:

• An effective and efficient spatiotemporal saliency model was developed to generate video saliency maps.
• A perceptual coding scheme was designed based on the generated saliency maps.
• A saliency-based quantization control algorithm was implemented to reduce bitrate.

Main Results:

• The proposed PVC framework achieved up to 9.46% bitrate reduction in objective and subjective tests.
• Negligible subjective and objective quality loss was observed compared to traditional methods.
• The method demonstrated superiority for high-definition video applications.

Conclusions:

• The novel PVC framework effectively reduces bitrate while preserving visual quality.
• The saliency-based approach offers significant advantages for video compression.
• This method is well-suited for high-definition video applications requiring efficient encoding.