This paper proposes a novel approach for offline text-independent writer identification. The proposed approach extracts two new features: Stroke Fragment Histogram (SFH) and Local Contour Pattern Histogram (LCPH). For SFH extraction, a handwriting image is firstly segmented into many stroke fragments (SFs) by using the proposed fragment segmentation method based on sliding window. Then all SFs extracted from training dataset are clustered to generate a codebook by using the Kohonen SOM 2D clustering algorithm. All SFs extracted from test datasets are adopted to compute SFHs by the proposed feature extraction method based on codebook. For LCPH extraction, the contour of an input handwriting image is firstly obtained Then a LCPH is formed to characterize the writer's individuality by tracking every contour point. For feature matching, the chi-square distance is employed to measure the similarity between SFHs and LCPHs. After feature matching, both similarities are fused for final decision by simple weighted sum. Three public handwriting datasets are used to evaluate the proposed approach and the experimental results show that the proposed approach can get the best performance compared with the state-of-the-art text-independent writer identification algorithms in all of these datasets.
28 February 2014 in International Journal of Signal Processing, Image Processing and Pattern Recognition
An image encryption algorithm based on hyperchaotic system is proposed to satisfy the security of image information transmission, the encryption algorithm consists of two times of scrambling and diffusion. The relation between plaintext and key is constructed to generate plaintext related chaotic sequence. Two rounds of scrambling operation are performed to make the pixel positions of the image are sufficiently scrambled, and two-way diffusion is carried out after scrambling, in this way, the value of any one pixel can affect other pixel values as much as possible. The information entropy, histogram distribution, adjacent pixel correlation, plaintext sensitivity, key sensitivity and key space are analyzed. As show by simulation results, our algorithm has good encryption effect and high security.
This paper proposes a novel saliency detection method by developing a deeply-supervised recurrent convolutional neural network (DSRCNN), which performs a full image-to-image saliency prediction. For saliency detection, the local, global, and contextual information of salient objects is important to obtain a high quality salient map. To achieve this goal, the DSRCNN is designed based on VGGNet-16. Firstly, the recurrent connections are incorporated into each convolutional layer, which can make the model more powerful for learning the contextual information. Secondly, side-output layers are added to conduct the deeply-supervised operation, which can make the model learn more discriminative and robust features by effecting the intermediate layers. Finally, all of the side-outputs are fused to integrate the local and global information to get the final saliency detection results. Therefore, the DSRCNN combines the advantages of recurrent convolutional neural networks and deeply-supervised nets. The DSRCNN model is tested on five benchmark datasets, and experimental results demonstrate that the proposed method significantly outperforms the state-of-the-art saliency detection approaches on all test datasets.