To overcome the limitation of unconstrained environment in face recognition, a modified algorithm using curvelet and DCNN (Madarkar and Sharma in J Intell Fuzzy Syst 38(5):6423–6435, 2020) is proposed. First, the curvelet transform of the face image is taken to obtain low frequency and high frequency components and DCNN are trained to make the image more robust to background and other changes. This results in a general feature vector and result is obtained by weighted fusion of all the DCNN. Finally the generalized CDCNN is used with random weights. The efficiency has been significantly improved by combining curvelet with DCNN for approximate band detailed sub band to improve the training and testing accuracy. DCNN is a well established technique for face recognition and is insensitive to small changes in input data. It possesses advantage of being ineffective to outliers and fast learning rate. The proposed method is robust-to-variation of imaging conditions. Performance comparison with other existing techniques shows that the proposed technique provides better results in terms of false acceptance rate, false rejection rate and recognition accuracy.

中文翻译：

---

在不受限制的环境下对来自互联网的视频进行人脸识别

为了克服人脸识别中不受约束的环境的局限性，提出了一种使用curvelet和DCNN的改进算法（Madarkar和Sharma in J Intell Fuzzy Syst 38（5）：6423-6435，2020）。首先，对人脸图像进行Curvelet变换以获得低频和高频分量，并对DCNN进行训练，以使图像对背景和其他变化更加鲁棒。这产生了通用特征向量，并且通过对所有DCNN进行加权融合获得了结果。最后，广义CDCNN用于随机权重。通过将Curvelet与DCNN结合用于近似频带详细子频带，可以显着提高效率，从而提高训练和测试的准确性。DCNN是一种成熟的面部识别技术，对输入数据的微小变化不敏感。它具有对异常值无效和学习速度快的优势。所提出的方法对成像条件具有鲁棒性。与其他现有技术的性能比较表明，所提出的技术在错误接受率，错误拒绝率和识别精度方面提供了更好的结果。