A Metric for Video Blending Quality Assessment

Area of Science:

• Computer Vision
• Digital Image Processing
• Video Editing Technology

Background:

• Video blending is crucial for smoothing intensity changes in video editing.
• Common blending techniques can introduce undesirable artifacts like bleeding and ghosting.
• Objective assessment of video blending quality is challenging.

Purpose of the Study:

• To propose an objective metric for assessing video blending quality.
• To incorporate both positive aspects (illuminance consistency) and negative aspects (artifacts) into the quality assessment.
• To evaluate the metric's performance against subjective evaluations and existing methods.

Main Methods:

• Developed a novel metric considering illuminance consistency and artifact detection (bleeding, ghosting).
• Incorporated temporal coherence between video frames into the metric.
• Evaluated the metric on a dedicated video blending dataset with subjective quality scores.

Main Results:

• The proposed metric effectively assesses video blending quality.
• Experimental results demonstrate superior performance compared to existing video quality metrics.
• The metric shows strong correlation with subjective quality evaluations.

Conclusions:

• The proposed objective metric provides a reliable and effective way to assess video blending quality.
• This approach offers a significant improvement over current video quality assessment methods.
• The metric can aid in developing better video blending algorithms and improving video editing workflows.