Face quality assessment has grown into a necessary pre-requisite for better performance of the face recognition (FR) algorithm pipeline. In this work, we proposed a novel face quality assessment algorithm for face recognition algorithms. We named it simplified face quality assessment (SFQA) for its simple provision to be used in accompaniment with any FR algorithm. We first proposed a hashing based deep learning model for the prediction of face quality from the features of the corresponding FR algorithm. The last layer of the deep hash net gives binary bits as output which is then converted to the decimal value to get the face quality score (FQS) ranging between 0 and 100. The mutual quality of the probe and gallery images are mathematically clubbed to get the Face Quality Confidence Score (FQCS). We have experimentally shown the effects of FQCS on the recognition algorithm. We trained the prediction model using the FR algorithm for the quality estimation on the face images from CAS-PEAL, LFW, and QDF database. The performance of SFQA is found to outperform the state-of-the-art methods, as discussed in the paper.

面部质量评估已成为提高面部识别（FR）算法管道性能的必要先决条件。在这项工作中，我们为面部识别算法提出了一种新颖的面部质量评估算法。我们将其命名为简化的面部质量评估（SFQA），是因为其简单的设置可与任何FR算法配合使用。我们首先提出了一种基于散列的深度学习模型，用于从相应的FR算法的特征预测面部质量。深层哈希网的最后一层提供二进制位作为输出，然后将其转换为十进制值，以得到介于0到100之间的面部质量得分（FQS）。人脸质量置信度得分（FQCS）。我们已经通过实验证明了FQCS对识别算法的影响。我们使用FR算法训练了预测模型，以对来自CAS-PEAL，LFW和QDF数据库的面部图像进行质量估计。如本文所述，发现SFQA的性能优于最新方法。