Detecting Images with Adult Content Using SURF and HaarWavelet

Abstract

Detecting images with adult content is one of the necessary and challenging problems in the fields of machine learning and machine vision. It can be used for a variety of applications such as content filtering and censoring, and user tracking, and banning. In this work, a framework for detecting nudity and adult content in digital images by using the SURF algorithm and Haar wavelet is developed. The proposed method is divided into online and offline modules. Each process that has a high time consumption is implemented in the offline module. Therefore, it can increase the speed of implementation of the model. For the proposed method, 899 images are used, in that, 402 images contain adult content, and 497 images have no adult content. 300 features of each image are extracted to create a bag of features (BOF), and they were used for the training by using the support vector machine (SVM). The results showed that the proposed method achieved an accuracy of 89.3%, a sensitivity of 86.2%, and a specificity of 92.03% for the classifier. The proposed framework provides an effective solution for detecting images and/or videos containing adult content from the internet environment.