Understandable Face Quality Assessment
Face recognition (FR) performance is affected significantly by the face image qualities, especially in real-world applications. Face image qualities vary significantly because of different imaging sensors, compression techniques, video frames, and/or image acquisition conditions/time. It is very challenging to assess face image qualities automatically, quickly and precisely in real world images.
Recent studies have shown that a learning-based paradigm can do better than the traditional heuristic methods, however, all these approaches can only give a single quality score as the “output,” e.g., 90, for an input face image. The “single-value quality score” cannot tell much information to communicate to human assessors. Further, many issues have not been addressed yet, e.g., what does a quality score mean? How to interpret a quality score with imaging conditions?
Why a face image has a quality score of 50 rather than 60? How well the quality scores characterize the real face image qualities? Can more useful cues (e.g., levels of details) be acquired to develop a complete representation for face image quality assessment?
In this project, we investigate a new paradigm, called understandable face image quality assessment, to address the issues in quality assessment of face images. We believe that the new paradigm can give a better solution for quality assessment.
The goal of this project is to develop a novel paradigm, which is understandable, believable, complete, and more accurate for face image quality assessment. It is also motivated by the “understandable template” in FBI, where some detailed information of faces is included during face template extraction. Our new quality representation has the potential to be integrated into the FBI’s understandable template. There are four tasks in this project.
- Task 1: Face and facial landmarks detection. Localizing faces in images is critical especially in real world images with cluttered background. Facial landmarks are essential to obtain facial details, e.g., distance between eyes, head pose, etc.
- Task 2: Explore measurements that are related to face image quality assessment and understandable to humans. We will develop a complete list of measures, e.g., distance between eyes, pose angle, in/out of focus, blurring levels, etc.
- Task 3: Improve quality scores based on the added measures. The measures developed in Task 2 will be used to improve or adjust the quality scores in an appropriate way. We study how to make the quality scores more accurate using the set of quality-related, understandable measures.
- Task 4: Validate and evaluate our approach experimentally.