Meet-in-the-middle: Multi-scale upsampling and matching for cross-resolution face recognition

TL;DR

This paper introduces a multi-scale approach combining super-resolution and matching techniques to improve cross-resolution face recognition, especially in surveillance scenarios, without requiring dataset-specific training.

Contribution

It presents a novel method that integrates face super-resolution, resolution matching, and multi-scale template accumulation for robust recognition across diverse image qualities.

Findings

Outperforms existing methods on surveillance face recognition benchmarks.

Does not require training or fine-tuning on target surveillance datasets.

Effective in recognizing faces from low-quality, long-range footage.

Abstract

In this paper, we aim to address the large domain gap between high-resolution face images, e.g., from professional portrait photography, and low-quality surveillance images, e.g., from security cameras. Establishing an identity match between disparate sources like this is a classical surveillance face identification scenario, which continues to be a challenging problem for modern face recognition techniques. To that end, we propose a method that combines face super-resolution, resolution matching, and multi-scale template accumulation to reliably recognize faces from long-range surveillance footage, including from low quality sources. The proposed approach does not require training or fine-tuning on the target dataset of real surveillance images. Extensive experiments show that our proposed method is able to outperform even existing methods fine-tuned to the SCFace dataset.