Face Recognition Under Dry and Wet Face Conditions

Research fraternities in face recognition were successful in addressing intraclass (intra-personal) and interclass (inter-personal) variations effectively. However, none of the works has made an attempt to address a special case of traditional face recognition, i.e., face recognition under dry and wet face conditions. Here, the gallery face is supposed to be dry face and the probe is a wet face, which comes into picture if the face recognition is employed for an automatic person authentication or in any other intelligent access control application. We name this scenario as wet face recognition (WFR). In such scenario, face remains wet due to several factors including adverse weather conditions such as rain, sweat, high humidity, snow fall, and fog. Essentially, face gets wet due to rain and sweat most commonly. Focus of the current work is to deal with wet image normalization and study its impact on face recognition rate. A framework based on modified bilateral filter is proposed to improve recognition performance in WFR scenario. Sparse representation-based classification over texture features results in impressive recognition performance. Extensive experiments on NITS-DWFDB database demonstrate the efficacy of the proposed wet normalization scheme and WFR framework.