Research on pornographic images recognition method based on visual words in a compressed domain

ORB feature based web pornographic image recognition

“…More than 4.6 million (ORB), 6.8 million (SIFT) keypoints are extracted and clustered to construct the visual vocabulary using K-means method respectively. The penalty parameter(C) of SVM error term is varied in a logarithmic scale from 2 -5 to 2 15 , and kernel parameter(g) is between 2 -15 and 2 3 [13] for selecting the optimal parameters by using the 5-fold cross-validation method and building the classifier model which can achieve the best recognition precision.…”

“…However, it's difficult to build a vast enough pre-labeled image dataset to cover all the diversity of images. The third category is to consider the pornographic image recognition as a classification problem (pornographic or non-pornographic images) [13,14] . Firstly these methods describe the contents of images by extracting some visual features (such as color, texture, outline, etc.…”

Social images tag ranking based on visual words in compressed domain

“…After the 128-D SIFT descriptor is obtained, some dimensionality reduction techniques are considered to speed the whole process. In this Section, Principal Component Analysis (PCA) is used for reducing the dimension of descriptors to 30 dimensions [25]. Suppose matrix X ¼ X 1 ; X 2 ; …; X N f g(N is the number of keypoints in one image) represents the SIFT descriptors and X i are column vectors, each of which has 128 rows.…”

“…The method utilized the spatial relationship of DCT coefficients between a block and its sub-blocks. In our previous work, much research about compressed domain has been explored [20][21][24][25][26]. Based on these studies, considering the compressed social images, social images tag ranking based on visual words in compressed domain is proposed in order to reduce the tag ranking time under ensuring the accuracy of social image tags.…”

Automatic classification of medical X‐ray images: hybrid generative‐discriminative approach

“…The BoWs model (sometimes also called bag of features or bag of visterns) represents images using histograms of quantised appearances of local patches. In recent years, many studies exploited this feature in various classification works [11][12][13][14][15][16][17][18][19][20][21][22][23]. Avni et al [12] proposed X-ray image categorisation and retrieval based on local patch representations using a 'bag of visual words' approach.…”