This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: Simultaneous Super-Resolution and Feature Extraction for Recognition of Low-Resolution Faces Pablo H. Hennings-Yeomans ECE Department Carnegie Mellon University [email protected] Simon Baker Microsoft Research Microsoft Corporation [email protected] B.V.K. Vijaya Kumar ECE Department Carnegie Mellon University [email protected] Abstract Face recognition degrades when faces are of very low resolution since many details about the difference between one person and another can only be captured in images of sufficient resolution. In this work, we propose a new pro- cedure for recognition of low-resolution faces, when there is a high-resolution training set available. Most previous super-resolution approaches are aimed at reconstruction, with recognition only as an after-thought. In contrast, in the proposed method, face features, as they would be extracted for a face recognition algorithm (e.g., eigenfaces, Fisher- faces, etc.), are included in a super-resolution method as prior information. This approach simultaneously provides measures of fit of the super-resolution result, from both reconstruction and recognition perspectives. This is dif- ferent from the conventional paradigms of matching in a low-resolution domain, or, alternatively, applying a super- resolution algorithm to a low-resolution face and then clas- sifying the super-resolution result. We show, for example, that recognition of faces of as low as 6 × 6 pixel size is con- siderably improved compared to matching using a super- resolution reconstruction followed by classification, and to matching with a low-resolution training set. 1. Introduction In many surveillance scenarios people may be far from the camera and the images of their faces may be small in the field of view. Such low resolution can seriously degrade the performance of conventional face recognition systems. Another similar scenario is when face recognition is used to help automatically organize family photographs, where often faces can be small. In this paper we study the problem of matching a low- resolution probe image to a high-resolution gallery of en- rolled faces. There are two standard approaches to this problem (see Figure 1 ): (1) Use super-resolution or inter- Matching 1 2 Probe Faces Gallery Templates Matching Matching S 2 R 2 Super-resolution Downsampling Figure 1. Standard approaches to matching a low resolution probe to a high resolution gallery. (1) Upsampling the probe (interpo- lation or super-resolution) and then matching. (2) Downsampling the gallery and then matching. In this paper we propose an alter- native algorithm that can outperform these two approaches. polation to reconstruct a higher resolution version of the low resolution probe and then perform matching in the usual way at higher resolution. (2) Downsample the entire gallery and then perform matching in low resolution....
View Full Document
This note was uploaded on 04/22/2010 for the course MI IP taught by Professor Vladbalan during the Spring '10 term at Universidad del Rosario.
- Spring '10