Adaptive Reservation of Network Resources According to Video Classification Scenes

Area of Science:

• Multimedia systems
• Signal processing
• Computer vision

Background:

• Effective video quality evaluation requires integrating subjective and objective metrics.
• Current methods often struggle to bridge the gap between Quality of Service (QoS) and Quality of Experience (QoE).
• Dynamic bitrate allocation is crucial for maintaining user satisfaction and efficient network resource utilization.

Purpose of the Study:

• To develop a novel approach for mapping QoS to QoE using QoE metrics.
• To establish a connection between objective and subjective video quality estimations.
• To create a predictive model for subjective video quality and optimize bitrate allocation.

Main Methods:

• Utilizing spatial information (SI) and temporal information (TI) from video frames to derive a sentinel flag for scene classification.
• Developing a comprehensive video database for quality evaluation, encompassing objective and subjective assessments.
• Employing an artificial neural network for subjective quality prediction based on objective evaluations and video characteristics (resolution, compression, bitstream).

Main Results:

• A new model accurately predicts subjective video quality using objective data and scene classification.
• An optimal mapping function was created to dynamically set variable bitrates based on scene type (sentinel flag).
• The model effectively balances user satisfaction, consistent video quality, and efficient network resource usage.

Conclusions:

• The proposed model enhances end-user comfort by ensuring consistent video quality and customer satisfaction.
• Dynamic bitrate allocation based on scene characteristics optimizes network resource utilization.
• The model provides accurate bitrate prediction for desired video quality, supporting both objective and subjective assessments.