基于人脸检测的网络图像内容过滤技术研究

摘要

为了有效防止不良图像的传播，文章提出了一种不良图片内容过滤方法。该方法通过建立一种 RGB、YUV 和 YIQ 空间上的混合肤色模型，实现裸露区域的检测。为有效降低正常图像的误判率，需对整个图像进行人脸检测。通过实验分析，该方法取得了较为理想的实验结果。%In order to effectively prevent the spread of sensitive images, this paper presents a method of content filtering of sensitive images. The method creates a RGB, YUV and YIQ skin color mixture model to detect exposed areas. In order to effectively reduce the error rate of normal images, face detection should be done for the whole picture. The experimental analysis shows that the method obtains a satisfactory result.