Liaoyuan Zeng

Feature Adaptive Co-Segmentation by Complexity Awareness

In this paper, we propose a novel feature adaptive co-segmentation method that can learn adaptive features of different image groups for accurate common objects segmentation. We also propose image complexity awareness for adaptive feature learning. In the proposed method, the original images are first ranked according to the image complexities that are measured by superpixel changing cue and object detection cue. Then, the unsupervised segments of the simple images are used to learn the adaptive features, which are achieved using an expectation-minimization algorithm combining l 1-regularized least squares optimization with the consideration of the confidence of the simple image segmentation accuracies and the fitness of the learned model. The error rate of the final co-segmentation is tested by the experiments on different image groups and verified to be lower than the existing state-of-the-art co-segmentation methods.

Semantic Annotation of Satellite Images Using Author–Genre–Topic Model

In this paper, we propose a novel hierarchical generative model, named author-genre-topic model (AGTM), to perform satellite image annotation. Different from the existing author-topic model in which each author and topic are associated with the multinomial distributions over topics and words, in AGTM, each genre, author, and topic are associated with the multinomial distributions over authors, topics, and words, respectively. The bias of the distribution of the authors with respect to the topics can be rectified by incorporating the distribution of the genres with respect to the authors. Therefore, the classification accuracy of documents is improved when the information of genre is introduced. By representing the images with several visual words, the AGTM can be used for satellite image annotation. The labels of classes and scenes of the images correspond to the authors and the genres of the documents, respectively. The labels of classes and scenes of test images can be estimated, and the accuracy of satellite image annotation is improved when the information of scenes is introduced in the training images. Experimental results demonstrate the good performance of the proposed method.

Low-Complexity Iterative Equalization for Symbol-Reconstruction-Based OFDM Receivers Over Doubly Selective Channels

Time selectivity of multipath channels introduces significant inter-carrier interference (ICI) in OFDM systems demanding high levels of mobility and capacity. In this paper, a piecewise channel approximation is presented for improving the structure of the frequency-domain channel gain matrix, leading to a low-complexity iterative equalization scheme for OFDM receivers. The equalization is a sequential data detection based on a parallel-filtering architecture, avoiding matrix inversion adopted by classical equalizers with ICI cancellation. Meanwhile, information redundancy in the traditional cyclically prefixed OFDM is investigated for time-invariant channels, then a method of symbol reconstruction (SR) is proposed to enhance OFDM system performance, including compensation for SNR loss and improvements on the accuracy of parameter estimation. On the basis of the appropriate channel modeling in this work, the SR processing is proved to be effective for OFDM receivers over doubly selective channels. The iterative equalization, combined with the SR processing, is applied to the DVB-H receiver design. Numerical simulation indicates that the proposed equalizer significantly improves the immunity of OFDM systems to time selectivity with very little complexity increased, in contrast to the conventional one-tap equalizer without ICI cancellation.

An effective vector model for global-contrast-based saliency detection

A basic vector model is derived to perform the global-contrast-based saliency detection.Two principles are provided to use the vector model effectively.The new vector model is developed by modifying the construction of the basic model.New model improves the detection precision, especially for large salient objects.Experimental results show the effectiveness of the proposed method. The saliency detection methods based on global contrast can generate full-resolution saliency map with uniformly highlighted regions and defined boundaries. For the images consisting of large salient objects, the use of unweighted sum of the color distances in the existing global-contrast-based methods may result in the detection of the background instead of the outstanding objects. In this paper, we propose a new global-contrast-based saliency detection method, called LRSW method, by deriving a new vector model which uses the weighted mean vector and contains the features of CIELAB color, chromatic double opponency, and similarity distribution. By using the vector model, the proposed method can significantly increase the detection precision and suppress the background in the saliency map, especially for large salient objects. The experimental results on the MSRA benchmark images show the effectiveness of the proposed method which outperforms the existing methods on visual saliency detection in terms of precision and recall.

A unified framework for spatiotemporal salient region detection

This article presents a new bottom-up framework for spatiotemporal salient region detection. The generated saliency map can uniformly highlight the salient regions. In the proposed framework, the spatial visual saliency and the temporal visual saliency are first computed, respectively, then they are fused with a dynamic scheme to generate the final spatiotemporal saliency map. In the spatial attention model, the approach of joint embedding of spatial and color cues is adopted to compute the spatial saliency map. In the temporal attention model, we propose a novel histogram of average optical flow to measure the motion contrast of the different pixels. The method can suppress the motion noise efficiently because the statistical distribution of optical flow in a patch is comparatively stable. Furthermore, we combine the spatial and the temporal saliency maps through an adaptive fusion method, in which a novel motion entropy is proposed to evaluate the motion contrast of the input video. Extensive experiments demonstrate that our method can obtain higher quality saliency map compared with state-of-the-art methods.

Adaptive Interpolation for Pilot-Aided Channel Estimator in OFDM System

In OFDM system, channel estimator with fixed-coefficient interpolation gives rise to significant estimation errors when the estimator mismatches the experienced time-varying channel. This paper presents a pilot-aided channel estimator adopting a double one-dimensional (1-D) polyphase subfilter structure, in which a few interpolating coefficients are applicable to all subcarriers of an OFDM symbol. By analyzing the channel interpolation for OFDM receiver, we propose a two-dimensional (2-D) adaptation scheme of the interpolating coefficient, based on a block least-mean-square algorithm, which requires none of knowledge on channel statistics. The 2-D adaptation scheme is simplified as two types of 1-D schemes respectively for the time and frequency dimensions, based on the known frequency- and time-direction interpolating coefficients. By simultaneously exploiting channel correlation in the time and frequency dimensions, the adaptive interpolators are trained to match channel properties. Furthermore, the proposed schemes of adaptive interpolation demand lower implementation cost, compared with the existing methods. Simulation results show that the adaptive interpolators converge at a high rate, and that the simplified schemes outperform the conventional methods in the high-speed mobile channel and the single-frequency-networks channel with long-delay echoes, while preserving a low implementation complexity.

Saliency-based superpixels

Superpixels provide an over-segmentation representation of a natural image. However, they lack information of the entire object. In this paper, we propose a method to obtain superpixels through a merging strategy based on the bottom-up saliency values of superpixels. The proposed method aims to obtain meaningful superpixels, i.e., make the objects as complete as possible. The proposed method creates an over-segmented representation of an image. The saliency value of each superpixel is calculated through a biologically plausible saliency model in a way of statistical theory. Two adjacent superpixels are merged if the merged superpixel is more salient than the unmerged ones. The merging process is performed in an iterative way. Experimental evaluation on test images shows that the obtained saliency-based superpixels can extract the salient objects more effectively than the existing methods.

Saliency detection in complex scenes

AbstractDetecting multiple salient objects in complex scenes is a challenging task. In this paper, we present a novel method to detect salient objects in images. The proposed method is based on the general ‘center-surround’ visual attention mechanism and the spatial frequency response of the human visual system (HVS). The saliency computation is performed in a statistical way. This method is modeled following three biologically inspired principles and compute saliency by two ‘scatter matrices’ which are used to measure the variability within and between two classes, i.e., the center and surrounding regions, respectively. In order to detect multiple salient objects of different sizes in a scene, the saliency of a pixel is estimated via its saliency support region which is defined as the most salient region centered at the pixel. Compliance with human perceptual characteristics enables the proposed method to detect salient objects in complex scenes and predict human fixations. Experimental results on three eye tracking datasets verify the effectiveness of the method and show that the proposed method outperforms the state-of-the-art methods on the visual saliency detection task.

Adaptive complex interpolator for channel estimation in pilot-aided OFDM system

In an orthogonal frequency division multiplexing system, conventional interpolation techniques cannot correctly balance performance and overhead when estimating dynamic long-delay channels in single frequency networks (SFNs). In this study, classical filter analysis and design methods are employed to derive a complex interpolator for maximizing the resistible echo delay in a channel estimator on the basis of the correlation between frequency domain interpolating and time domain windowing. The coefficient computation of the complex interpolator requires a key parameter, i.e., channel length, which is obtained in the frequency domain with a tentative estimation scheme having low implementation complexity. The proposed complex adaptive interpolator is verified in a simulated digital video broadcasting for terrestrial/handheld receiver. The simulation results indicate that the designed channel estimator can not only handle SFN echoes with more than 200 μs delay but also achieve a bit-error rate performance close to the optimum minimum mean square error method, which significantly outperforms conventional channel estimation methods, while preserving a low implementation cost in a short-delay channel.

Saliency detection using a central stimuli sensitivity based model

In this paper, a novel method is proposed to predict attention in image scenes by using a central stimuli sensitivity based saliency model. The proposed method is based on the general “center-surround” visual attention mechanism and the spatial frequency response of the human visual system (HVS). Following three biologically inspired principles, the saliency value is computed by two “scatter matrices” which are used to measure the similarity and distinctness within and between two classes, i.e., the center and surrounding regions, respectively. In order to detect salient objects with different size, the saliency of a pixel is estimated via the saliency support region of the pixel, which is the most salient region centered at the pixel with respect to the surrounding region. The proposed method which is compliant with human perceptual characteristics enables the prediction of human fixations. Experimental results on three eye tracking datasets verify the effectiveness of the method and show that the proposed method outperforms the state-of-the-art methods on the visual saliency detection task.