人脸图像补全的方法用途及研究

Research and Application of Face Image Completion Method

摘要: 目前，监控视频在侦查破案中起到了关键性作用，同时随着深度学习为代表的人工智能的迅速发展，利用人脸进行身份识别已成为主流破案手段。然而，由于犯罪嫌疑人故意遮挡或视频分辨率等原因，造成人脸关键部位信息丢失，使人脸识别系统不能够完成人脸匹配。因此，本文拟针对上述问题，对遮挡的、模糊的人脸图像的处理方法开展相关研究。通过对图像的平滑去噪，去除图像的尖锐噪声，消除噪声的干扰；瞳孔检测精确定位眼睛位置，将特征向量归一化，以此来解决人脸尺度变化和人脸旋转问题；最后移除人脸图像的障碍物，利用领域信息填充缺损的人脸区域达到补全的效果。

Abstract: At present, surveillance video plays a key role in investigating and solving cases. At the same time, with the rapid development of artificial intelligence represented by in-depth learning, face recog-nition has become a mainstream means of solving cases. However, due to the intentional occlusion or video resolution of the suspect, the information of key parts of the face is lost, which makes the face recognition system unable to complete face matching. Therefore, in view of the above problems, this paper intends to carry out relevant research on the processing methods of occluded and blurred face images. By smoothing and denoising the image, eliminating the sharp noise of the image, eliminating the noise interference; pupil detection accurately locating the eye position, normalizing the feature vectors, the problems of face scale change and face rotation are solved. Finally after removing the obstacles of the face image, the defective face area is filled with the do-main information to achieve the effect of completing.