Radial Distortion in Face Images: Detection and Impact

Wassim Kabbani, Tristan Le Pessot*, Kiran Raja, Raghavendra Ramachandra, Christoph Busch, ENSICAEN, Caen, France* · 2025

Read Radial Distortion in Face Images: Detection and Impact by Wassim Kabbani, Tristan Le Pessot*, Kiran Raja, Raghavendra Ramachandra, Christoph Busch, ENSICAEN, Caen, France* (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-07179.txt.

Introduction

Wassim Kabbani, Tristan Le Pessot, Kiran Raja, Raghavendra Ramachandra, and Christoph Busch build a detector for fish-eye warp in face photos used for ID checks. Their paper from twenty twenty-five targets enrolment quality in face recognition systems. Radial distortion bends facial traits and drags down match scores. Self-enrolment by phone makes such flaws easy to miss.

Their model flags warped enrolment shots before they enter the system. It is cast as a face image quality algorithm, in line with field practice. The team also gauges how warp size shifts recognition scores in depth. Detection marks came out strong in their trials. The impact study yields rules for how flagged shots should be handled in live systems. The work guards the front door of identity pipelines.

Catalog record

CREATORWassim Kabbani, Tristan Le Pessot*, Kiran Raja, Raghavendra Ramachandra, Christoph Busch, ENSICAEN, Caen, France*
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-07179

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Radial Distortion in Face Images: Detection and Impact Wassim Kabbani, Tristan Le Pessot*, Kiran Raja, Raghavendra Ramachandra, Christoph Busch, ENSICAEN, Caen, France* Abstract Acquiring face images of sufficiently high quality is important for online ID and travel document issuance applications using face recognition systems (FRS). Low-quality, manipulated (intentionally or unintentionally), or distorted images degrade the FRS performance and facilitate documents’ misuse. Securing quality for enrolment images, especially in the unsupervised self-enrolment scenario via a smartphone, becomes important to assure FRS performance. In this work, we focus on the less studied area of radial distortion (a.k.a., the fish-eye effect) in face images and its impact on FRS performance. We introduce an effective radial distortion detection model that can detect and flag radial distortion in the enrolment scenario. We formalize the detection model as a face image quality assessment (FIQA) algorithm and provide a careful inspection of the effect of radial distortion on FRS performance. Evaluation results show excellent detection results for the proposed models, and the study on the impact on FRS uncovers valuable insights into how to best use these models in operational systems. 1 Introduction (a) Original (b) DM \lambda=0.4 (c) DM \lambda=0.9 (d) Original (e) DM \lambda=0.4 (f) DM \lambda=0.9 Figure 1 : Radial distortion effect in face images. The left column shows the original, undistorted images. The middle column shows images with a mild level of distortion. The right column shows the images with a more pronounced level of distortion. Distortion is introduced synthetically using the Division Model (DM) but with different values for the distortion coefficient \lambda . Images from the FRLL dataset DeBruine2021 . Face recognition systems (FRS) are being deployed at a number of airports, border crossings, and security gates. They have become a cornerstone of our security architecture to verify identity. Online ID and travel document issuance applications that are mobile/smartphone-compatible are being rolled out in many countries 1 1 1 https://www.ireland.ie/en/dfa/passports/passport-online/ 2 2 2 https://www.cbp.gov/travel/us-citizens/mobile-passport-control . These applications allow individuals to self-enroll by uploading an existing face photo or taking a selfie, which will then be used on the issued document as a reference image to verify the identity of the individual in a later encounter. The reliability of these systems and the reference data stored therein is therefore of great importance. FRS needs to not only perform well but also be quality-aware and immune to various kinds of attacks. Many studies have shown that low-quality images have a negative impact on FRS performance ISO-IEC-29794-5-DIS-FaceQuality-240129 ; Schlett-FIQA-LiteratureSurvey-CSUR-2021 ; Boutros-CR-FIQA-CVPR-2023 , and further, they have a negative impact on detecting presentation attacks (also morphing) Aravena-2022 ; Biying-2022 . Robust methods that can examine the quality of the incoming images are therefore needed to flag any non-complaint or suspected image, making sure the obtained images are as representative as possible of the individual to whom they belong. According to the ISO/IEC 29794-5 standard ISO-IEC-29794-5-DIS-FaceQuality-240129 , the quality assessment of a face image belongs to capture-related and subject-related quality components. Capture-related quality components such as background uniformity, illumination uniformity, under/over exposure, and natural color, and subject-related quality components such as eyes open, mouth closed, head size, and pose have dedicated metrics to measure them. Radial distortion is yet another capture-related quality component emerging due to the optical properties of the camera. Commonly known as the fish-eye effect, it distorts the captured image, creating a hemispherical (panoramic) effect. The disturbance effect on the image and the perception of the represented face are shown in the example images in Figure 1 . Such an image distortion is an effect caused by user interaction that is not compliant with the capture process policy (for example, the smartphone is too close to the subject) and is not necessarily an attack. The standard ISO/IEC 19794-5 ISO-IEC-19794-5-G2-FaceImage-110304 mandates the ”fish eye” effect associated with wide-angle lenses to be absent. Further, the standard ISO/IEC 39794-5 ISO-IEC-39794-5-G3-FaceImage-191015 requires the maximum magnification distortion rate to be equal to or below 7%. However, to the best of our knowledge, an algorithm to detect or estimate radial distortion in a face image is not yet contained in the draft quality testing version of the implementation-focused standard ISO/IEC 29794-5 ISO-IEC-29794-5-DIS-FaceQuality-240129 . In the course of addressing the above-mentioned limitation, this paper primarily contributes with: (1) An effective radial distortion detection model that fits the unsupervised self-enrolment scenario and can flag radially distorted face images with high confidence. (2) A face image quality assessment (FIQA) algorithm based on the detection model that can produce a quality measure for the radial distortion quality component. (3) Insights into when to best run checks, such as radial distortion in an operational scenario. The secondary contributions are: (1) Formalizing the distortion transformations in a unified way, making them easier to use in future works. (2) Synthetic datasets for training and evaluating radial distortion models 3 3 3 The code and datasets will be made public along with the paper. . The paper is organized as follows: Section 2 presents related work and preliminaries. Section 3 introduces the proposed approach. Section 4 presents experiments and results. Sections 5 and 6 present discussions and conclusions. 2 Related Works 2.1 Face Image Quality Assessment (FIQA) Face Image Quality Assessment (FIQA) refers to the process of assessing the quality of a face image and its utility for a face recognition system Schlett-FIQA-LiteratureSurvey-CSUR-2021 . As there are many aspects of the face image that could contribute to its quality, these aspects are being standardized in the ISO/IEC 29794-5 standard ISO-IEC-29794-5-DIS-FaceQuality-240129 , which uses the standardized term measure to refer to all quality aspects of a face image. A FIQ measure can be either an end-to-end unified quality score that gives an overall assessment of the image or a quality component that assesses the quality of a specific aspect (or potentially a defect) of the face image, such as background uniformity, illumination uniformity, and natural color. 2.1.1 Unified Quality Score Many end-to-end methods that produce a unified quality score have been introduced. Hernandez-Ortega et al . introduced FaceQNet V0 hernandezortega2019faceqnet and an optimized version later on in 2021 with FaceQNet V1 hernandezortega2021biometric . Boutros et al . Boutros-CR-FIQA-CVPR-2023 presented a method for unified FIQA that depends on measuring the relative sample classifiability, which is measured based on the allocation of the training sample feature representation in angular space with respect to its class center and the nearest negative class center. They illustrate that there is a correlation between the face image quality and the sample’s relative classifiability. 2.1.2 Quality Components Several works addressed individual quality components. Hernandez-Ortega hernandez-ortega2021FaceQvec proposed a framework to assess several quality components such as exposure, unnatural color, expression, pose, and background uniformity. The framework assesses image exposure by detecting if the image is too dark (or too bright) based on the mean pixel value. It uses a color-detector model to detect pixels with unnatural color and a pretrained CNN to detect non-neutral facial expression. Grimmer et al . NeutrEx-IJCB-2023 proposed an expression neutrality quality assessment method, NeutrEx , that is based on the accumulated distances of a 3D face reconstruction to a neutral expression anchor EMOCA:CVPR:2021 DECA:Siggraph2021 . Yet, no work has addressed radial distortion in facial images and its utility as a face

Read the complete text