Why Video Quality Matters and How Lingjing Revolutionizes Its Evaluation

Introduction

Video services need to compare different videos, encoding parameters, and products to determine which offers the best visual experience for users.

Factors Affecting Video Quality

Resolution – higher pixel count yields finer detail.

Frame rate – number of images per second (e.g., 24 fps for movies, 30 fps for TV).

Brightness – perceptible range from 10⁻³ nits to 10⁶ nits.

Bit depth – more bits per pixel allow smoother color gradients.

Color gamut – represented by the area on the CIE 1931 diagram.

Bitrate – higher bitrate can improve quality but may waste bandwidth.

Why Improve Quality?

Higher quality can enhance user experience and product metrics, yet the actual impact on watch time, content influence, or user preference is not always clear. Device capabilities, bandwidth costs, and competitive pressure also drive the need for better quality.

Importance of Quality Evaluation

Accurate measurement is essential for improving codecs, setting system requirements, and benchmarking against competitors. Both subjective (user opinion) and objective (algorithmic) assessments are needed, but each has limitations.

Subjective Evaluation

Time‑consuming and not scalable.

Results vary across individuals.

Objective Evaluation

Algorithms such as VMAF, PSNR, and SSIM model subjective scores, providing fast, repeatable measurements, yet they must stay aligned with human perception.

Factors Influencing Evaluation Confidence

Confidence can be compromised by sample selection (duration, scene variety, logo presence), tool limitations (scaling errors, overlapping displays, motion masking), and evaluator bias (order effects, focus differences).

Lingjing: A Comprehensive Evaluation System

Developed over two years with 500+ subjective tests, 1,000+ samples, and thousands of evaluators, Lingjing follows ITU standards and incorporates more than ten patented technologies. It supports multi‑platform (PC, Android, iOS) and multiple testing modes, including single‑screen, overlapped‑screen (DCR/CCR), expert, and side‑by‑side evaluations.

Key capabilities:

Camera mock – replace camera input with specified video.

Objective metrics – VMAF, PSNR (with spatial visualization), SSIM.

Video attribute extraction – SITI, chroma, saturation, brightness, bitrate, frame rate, resolution.

Video processing & detection – static/frame detection, audio‑video sync, anomaly generation.

The system is built on a layered architecture (infrastructure, tool, service) to ensure flexibility and business alignment.

Current Deployments and Future Roadmap

Lingjing is used by Baidu Feed, Haokan, live streaming, cloud storage, and Xiaodu services. Planned enhancements include single‑frame comparison, HDR evaluation, 8K / 120 fps testing, and specialized pipelines for beauty and makeup content.