Face Photo-Sketch Transformation and Population Generation

摘要

The problem of the automatic recognition and processing of face photos and sketches is of significant importance for law enforcement applications. A photo-sketch transformation is a convenient way to reduce dissimilarities between a face sketch and a photo before employing one of the existing face recognition techniques. In this paper, we propose to generate a photo (sketch) population instead of a single image. This approach improves face recognition rate and reduces the effect of different appearance variations. The proposed method is based on the modified version of the Eigenface/Eigensketch approach. We carried out the experiments on the photo-sketch transformation and population generation. The obtained results confirm the capability of the proposed technique to transform sketches of various styles to photos and introduce an inter-population diversity that is sufficient to cover various face appearance deviations.