How VQML® Automates Human-Perceived Video Quality Assessment Beyond Network KPIs

For decades, video quality assessment has relied heavily on subjective testing.

Large evaluation panels—sometimes involving hundreds of participants—were assembled to use mobile video services, watch the delivered content, and rate its visual quality on a five-point scale: 1 Bad · 2 Poor · 3 Fair · 4 Good · 5 Excellent

The average of these individual ratings was calculated as the Mean Opinion Score (MOS), providing a benchmark based directly on human perception. Although subjective testing remains valuable, it requires considerable time, cost, preparation, and operational effort. More importantly, it cannot continuously assess the vast number of real-time video sessions generated across today’s mobile networks, streaming platforms, conferencing services, and surveillance systems.

The industry therefore needs a new approach—one that can reproduce human visual judgment automatically, objectively, and at scale.

Why Video Quality Assessment Must Change

Network KPIs such as RSRP, SINR, throughput, latency, and packet loss remain essential for understanding network conditions. However, they do not directly answer the question that matters most to the viewer:

Was the video actually clear, stable, and usable?

A network may report sufficient throughput and stable radio conditions while users still experience macroblocking, blur, freezing, resolution degradation, or unstable video calls. Video quality can also be affected by factors outside the network, including compression, camera shake, poor focus, low-light noise, device resolution, rendering performance, and application behavior.

As networks evolve toward 5G Standalone and future 6G architectures, quality assurance must move beyond infrastructure performance and measure the visual experience actually delivered to the user. With the rapid advancement of AI, ITU-T has now established an objective and rigorous framework for standardizing AI-based, Non-Refence video quality assessment.

What Is VQML®? From AI Innovation to an International Standard

VQML® is LIG Accuver’s proprietary AI-based No-Reference video quality assessment technology. Powered by an AI-driven assessment engine, VQML® predicts human-perceived video quality directly from the decoded RGB frames received by the user. It does not require:

The original reference video Transmission or encoding metadata Modification of the target application Repeated subjective testing panels

VQML® analyzes received video and generates a predicted MOS on a scale from 1 to 5, either in real time or at the session level.

Developed independently by LIG Accuver, VQML® was selected as Model A under the J.noref standardization initiative at an ITU-T Study Group 12 plenary meeting in Geneva. The initiative subsequently matured into the approved ITU-T J.344 recommendation family for No-Reference objective video quality assessment of Full HD video.

Its capabilities cover both major areas of the standard:

ITU-T J.344.1: Coding artifacts caused by video compression and processing ITU-T J.344.2: Combined coding and camera-related impairments, including blur, camera shake, focus errors, sensor noise, and lighting-related degradation

VQML® is the only registered model applied across both J.344.1 and J.344.2, and the sole registered model within J.344.2. This establishes an internationally recognized technical foundation for AI-based perceptual video quality measurement.

Evaluation using the ITU-T standardization dataset further demonstrated VQML®’s strong alignment with subjective human quality scores. It achieved a Pearson Linear Correlation Coefficient of 0.9041. On a dataset containing severe coding artifacts, VQML® also recorded an RMSE of 0.4384 and an FRR of 4.01%, demonstrating low prediction error and reliable quality-ranking consistency under the evaluated conditions.

These results reflect not only the accuracy of the model, but also its ability to interpret video quality in a way that more closely aligns with human perception.

Where VQML® Creates Value

VQML® has undergone commercial validation in disaster-network verification projects associated with Korea’s public-safety communications environment, as well as evaluations with global mobile operators and network equipment vendors.

Its representative applications include:

Mobile Video and Short-Form Benchmarking: Operators can objectively compare video quality across YouTube, TikTok, Instagram, OTT platforms, devices, networks, and mobility conditions. Video Conferencing Validation: Services such as Zoom, Microsoft Teams, and Webex can be evaluated for facial clarity, motion continuity, macroblocking, and session stability. Mission-Critical Communications: VQML® can help verify whether live video remains sufficiently clear and usable during congestion, mobility, and degraded radio conditions. CCTV and Smart-City Monitoring: Large-scale camera environments can be screened for focus drift, camera vibration, low-light noise, sensor defects, and compression-related degradation. Laboratory and Device Validation: XCAL’s Virtual Camera function can inject standardized video into a device, transmit it through a test network, and evaluate the received output under controlled conditions. External Playback Device Testing: Video from Apple TV, Chromecast, set-top boxes, and comparable streaming devices can be evaluated through an external HDMI capture interface without requiring access to the original source.

From Standardization to Commercial Deployment

VQML® is planned for deployment across GPU Server, PC, and Mobile environments, supporting applications from centralized high-volume processing to laboratory, field, and device-side measurement. Within XCAL, VQML® can assess mobile video services directly or evaluate content from set-top boxes and streaming devices through an external capture interface.

LIG Accuver also plans to offer VQML® as a commercially licensed SDK, enabling customers to integrate the J.344-aligned engine into their own validation, monitoring, and service-assurance platforms. A patent application for the core technology further supports future licensing and intellectual-property-based commercialization.

👉 Read more about VQML® White Paper

Let’s connect and build the future together!

📬 For direct inquiries or to discuss your testing needs, contact us here.

📩 Want insights delivered to your inbox? Subscribe to our newsletter for the latest in wireless testing, product innovation, and more.