He Zifen

Unsupervised image sequence segmentation based on hidden Markov tree model

This paper presents a novel unsupervised image sequence segmentation method using hierarchical wavelet domain hidden Markov tree model(WDHMT). The key idea is that with a priori information introduced into the segmentation framework, we can capture both local and global statistical information using WDHMT model. Firstly, each frame extracted from the image sequence is transformed through discrete wavelet transform(DWT) to obtain a compressive representation of the original one. Then we capture the context information of wavelet coefficients at each level through tree-structured probabilistic graph. After the model parameters are learned through up-down iterated expectation maximization(EM) algorithm, we deduced the maximum likelihood(ML) segmentation at the finest level. The boundary information is then fused with the a priori region information. Finally, we quantitatively evaluated the performance of this algorithm by using a sequence of tobacco leaf images polluted by Gaussian white noise. The simulation results show that the proposed algorithm can achieve high classification accuracy, preferable specificity and sensitivity properties.

Detecting System of Ink Cells in Gravure Cylinder via Neural Network

We apply neural network to build up a detecting system of ink cells in gravure cylinder. Firstly, the ink cells images are gained in the images capturing device and histogram equalization. The edge of cells is extracted by using of Canny operator. We use different thresholds and experimental sigma values that compare to experimental results. Canny edge extraction operator is best when the value of sigma is 16. According to the image used in this research to determine the standard ink cells carving the value of gaps d0 equals 125, the value of dark tone s0 equals 394, so its standard value of gaps and dark tone are d0 ± 10 and s0 ± 10. The value of gravure outlets gaps and dark tone are measured, while d and s is in the scope of standard range, which the output 1 of the ink cells determined to pass and the output 0 deemed to fail. Binarization images are obtained through adaptive threshold segmentation, which regards the value of gaps and dark tone as the characteristic value when they start to detecting. Finally, we extract size and surface defects of ink cells for grading. Segmentation pictures are extracted by K-means clustering. The areas of ink cells are deemed to size characteristics. Then we classify the ink cells into two classes by using of neural network. The experimental results consider neural network model that produce consequences.

Digital Watermarking Algorithm Based on WLSMB Halftoning Image

In this paper, a novel digital watermarking scheme for halftone image is proposed. Halftone images are obtained by used of the weighted least-squares model-based(WLSMB) and cluster method. We applied to embed digital watermark to protect the suffering during transmission compression attacks and geometric attacks robustness. Experimental results show that the presented halftone image watermarking is invisible and robust against various signal processions such as JBIG compression, noise adding, cropping, daubing and print-scan.

Single Loop Inference of Hidden Markov Tree for Multiscale Image Segmentation

This paper addresses the problem of bottom-up and up-bottom multiscale segmentation of objects in the presence of dynamic backgrounds. Previous hidden Markov tree (HMT) based approaches have exploited an iterative inference scheme and each iteration consists of an two-stage segmentation mechanism, namely, parameter learning and multiscale fusion of likelihoods. In this paper, we propose a novel approach for recovering multiscale segmentation accurately in the absence of iterative multiscale fusion stage. This allows both inference and fusion of multiscale classification likelihoods to be computed in a single loop through bottom-up likelihood estimation and up-bottom posterior inference of HMT. Experimental results on a synthesized image in the presence of Gaussian white noise demonstrate the high robustness achieved by the proposed method.