[论文解读] Surveillance Facial Image Quality Assessment: A Multi-dimensional Dataset and Lightweight Model

Yanwei Jiang, Wei Sun+9 more|arXiv (Cornell University)|Feb 7, 2026

Image and Video Quality Assessment被引用 0

一句话总结

Introduces SFIQA-Bench, a multi-dimensional surveillance facial image quality benchmark, and SFIQA-Assessor, a lightweight multi-task model that jointly assesses perceptual quality and fidelity across six dimensions. The approach outperforms state-of-the-art FIQA/IQA methods with real-time efficiency on surveillance data.

ABSTRACT

Surveillance facial images are often captured under unconstrained conditions, resulting in severe quality degradation due to factors such as low resolution, motion blur, occlusion, and poor lighting. Although recent face restoration techniques applied to surveillance cameras can significantly enhance visual quality, they often compromise fidelity (i.e., identity-preserving features), which directly conflicts with the primary objective of surveillance images -- reliable identity verification. Existing facial image quality assessment (FIQA) predominantly focus on either visual quality or recognition-oriented evaluation, thereby failing to jointly address visual quality and fidelity, which are critical for surveillance applications. To bridge this gap, we propose the first comprehensive study on surveillance facial image quality assessment (SFIQA), targeting the unique challenges inherent to surveillance scenarios. Specifically, we first construct SFIQA-Bench, a multi-dimensional quality assessment benchmark for surveillance facial images, which consists of 5,004 surveillance facial images captured by three widely deployed surveillance cameras in real-world scenarios. A subjective experiment is conducted to collect six dimensional quality ratings, including noise, sharpness, colorfulness, contrast, fidelity and overall quality, covering the key aspects of SFIQA. Furthermore, we propose SFIQA-Assessor, a lightweight multi-task FIQA model that jointly exploits complementary facial views through cross-view feature interaction, and employs learnable task tokens to guide the unified regression of multiple quality dimensions. The experiment results on the proposed dataset show that our method achieves the best performance compared with the state-of-the-art general image quality assessment (IQA) and FIQA methods, validating its effectiveness for real-world surveillance applications.

研究动机与目标

• 定义在监控图像中评估感知质量和人脸保真度的挑战。
• 创建 SFIQA-Bench，这是一个具有真实世界监控图像和六个质量维度的多维基准。
• 提出 SFIQA-Assessor，这是一个轻量级模型，使用多视角面部输入和跨视角注意力进行多任务质量回归。
• 证明 SFIQA-Assessor 在六个质量维度上实现了与最先进 FIQA/IQA 方法相当的性能，同时具备实时效率。

提出的方法

• 用三种摄像机模型在室内、室外和 ITS 车辆场景中构建含 5,004 张监控人脸图像的 SFIQA-Bench。
• 从 100 名参与者处在六个维度上收集主观评分：噪声、清晰度、色彩丰富度、对比度、保真度和整体质量。
• 将 SFIQA-Assessor 设计为一个轻量级多任务 FIQA 模型，使用三个面部视图（原始、面部、眼睛与口部）作为输入。
• 使用面部感知的质量特征编码器提取多尺度特征，并使用跨视图注意力模块融合视图。
• 实现一个具有任务感知解码器的模型，包含任务自注意力和跨注意力，随后为六个质量分数设置单独的回归头。
• 通过面部检测与面部解析对输入进行预处理，生成三种视图，从而在面部区域实现聚焦质量评估。

实验结果

研究问题

• RQ1 如何在监控图像中联合评估感知质量和人脸保真度？
• RQ2 一个多视图、轻量级模型能否实时准确预测多维质量维度？
• RQ3 在 SFIs 中，各质量因素（噪声、清晰度、色彩丰富度、对比度、保真度）对整体感知质量的相对重要性如何？
• RQ4 环境条件（日夜、室内/室外、车辆/行人）是否影响监控数据中的多维质量判断？

主要发现

• SFIQA-Bench 为 SFIs 提供六个人类质量维度，覆盖三个监控场景共 5,004 张图像。
• MOS 分析显示清晰度和保真度是对总体质量的主要贡献者，日间场景通常比夜间评分更高。
• 提出的 SFIQA-Assessor 在六个维度上均优于最先进的一般 IQA/FIQA 方法，同时保持较低的计算成本。
• 回归模型显示保真度、清晰度和对比度与总体质量高度一致，支持多维评估方法。
• 跨视图融合和面向任务的解码使从三种面部视图实现有效的多任务质量预测成为可能。