Visual Quality Evaluation of Image Object Segmentation: Subjective Assessment and Objective Measure

Area of Science:

• Computer Vision
• Image Processing
• Human-Computer Interaction

Background:

• Visual quality evaluation is crucial for image object segmentation.
• Developing objective measures that correlate with human perception is an ongoing research challenge.
• A standardized platform is needed to assess the performance of these objective measures.

Purpose of the Study:

• To introduce a novel subjective database for object segmentation visual quality assessment.
• To propose a new full-reference objective measure for evaluating object segmentation quality.
• To validate the proposed measure against existing state-of-the-art methods.

Main Methods:

• Creation of a subjective database with 255 object segmentation results, assessed by over 30 human subjects.
• Development of a novel full-reference objective measure incorporating four human visual properties.
• Comparative analysis of the proposed measure with existing objective measures using the created database.

Main Results:

• The proposed objective measure demonstrates superior performance in matching subjective human judgments compared to state-of-the-art methods.
• The experimental results validate the effectiveness of the novel objective measure.
• The developed subjective database is publicly released for community use.

Conclusions:

• The novel objective measure provides a more accurate assessment of object segmentation visual quality.
• The public database facilitates further research and development in objective visual quality evaluation.
• This work contributes to advancing the field of image object segmentation evaluation.