Evaluating Quality of Screen Content Images Via Structural Variation Analysis.

Area of Science:

• Computer Vision
• Image Processing
• Human-Computer Interaction

Background:

• Computer-generated signals, particularly screen content images, are prevalent in daily life.
• Screen content images present unique challenges for research areas like compression, transmission, and quality assessment due to their graphic and textual components.
• Existing research has primarily focused on natural scene images, leaving screen content image analysis less explored.

Purpose of the Study:

• To develop a novel quality evaluation model for screen content images.
• To address the challenges posed by graphic and textual elements in screen content images.
• To improve the accuracy of screen content image quality assessment.

Main Methods:

• Analyzing structural variations in screen content images, categorizing structures into global and local.
• Considering characteristics of graphic and textual images, such as limited color variations.
• Incorporating principles of the human visual system into the quality assessment model.
• Systematically combining measurements of global and local structural variations for final quality estimation.

Main Results:

• The proposed quality model was evaluated on three screen content image quality databases.
• Experimental results demonstrated the effectiveness of the model in assessing image quality degraded by capturing, compression, and transmission.
• The model showed superior performance compared to state-of-the-art relevant methods.

Conclusions:

• The developed quality evaluation model provides a robust approach for assessing screen content images.
• The method effectively captures quality degradation by analyzing structural variations.
• This work offers a significant advancement in the field of screen content image quality assessment.