Restoration of motion-blurred image based on border deformation detection: a traffic sign restoration model.

Zeng, Yiliang; Lan, Jinhui; Ran, Bin; Wang, Qi; Gao, Jing · 2015 · DOAJ

DOI: 10.1371/journal.pone.0120885

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of restoring motion-blurred traffic sign images for Driver Assistance Systems (DAS). As vehicles move, relative motion between the camera and the scene causes motion blur, which degrades image quality and hinders accurate traffic sign recognition. While hardware improvements have limitations, software-based restoration algorithms are essential. Existing methods often rely on frequency-domain analysis (e.g., spectral zero patterns) or blind deconvolution, which can be computationally expensive or sensitive to noise. The authors propose a novel spatial-domain algorithm based on "border deformation detection" to quickly and accurately estimate motion parameters for restoration. The method exploits the geometric properties of traffic signs, which typically feature distinct circular borders. The algorithm first converts the image from RGB to HSI color space to extract the red border using hue thresholds, a process robust to lighting variations. It then calculates the center of the border and measures the border width in all directions (0° to 180°). The core insight is that motion blur causes the border to widen along the direction of motion while remaining relatively unchanged perpendicular to it. By identifying the maximum and minimum border widths, the algorithm determines the motion direction (aligned with the minimum width) and scale (derived from the difference between maximum and minimum widths, adjusted by a correction coefficient). These parameters define the Point Spread Function (PSF), which is then used in a Wiener filter to restore the image. Experiments were conducted using the German Traffic Sign Recognition Benchmark (GTSRB) dataset, which contains over 50,000 images across more than 40 classes. The proposed method was compared against traditional blind deconvolution and Lucy-Richardson methods. The results demonstrated that the border deformation detection approach effectively restores motion-blurred images with lower computational cost. The authors introduced a Gray Mean Grads (GMG) ratio to evaluate restoration quality, showing that their method significantly improved image clarity. Consequently, the restored images led to higher correct recognition rates for traffic signs compared to the baseline methods. The significance of this work lies in providing an efficient, spatial-domain solution for a critical problem in intelligent transportation. By focusing on the specific structural features of traffic signs (the border), the method avoids the complexity and noise sensitivity of frequency-domain approaches. This enables real-time processing capabilities suitable for DAS, enhancing driver safety by ensuring reliable sign recognition even under motion blur conditions. The study confirms that leveraging prior knowledge of object geometry can yield robust and computationally efficient image restoration results.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• sign visibility legibility