Detection of salient objects from images is gaining increasing research interest in recent years as it can substantially facilitate a wide range of content-based multimedia applications. Based on the assumption that foreground salient regions are distinctive within a certain context, most conventional approaches rely on a number of hand-designed features and their distinctiveness is measured using local or global contrast. Although these approaches have been shown to be effective in dealing with simple images, their limited capability may cause difficulties when dealing with more complicated images. This paper proposes a novel framework for saliency detection by first modeling the background and then separating salient objects from the background. We develop stacked denoising autoencoders with deep learning architectures to model the background where latent patterns are explored and more powerful representations of data are learned in an unsupervised and bottom-up manner. Afterward, we formulate the separation of salient objects from the background as a problem of measuring reconstruction residuals of deep autoencoders. Comprehensive evaluations of three benchmark datasets and comparisons with nine state-of-the-art algorithms demonstrate the superiority of this paper.

/pdf/background-prior-based-salient-object-detection-via-deep-4rebjv1qs9.pdf

Background Prior-Based Salient Object Detection via Deep Reconstruction Residual

The motion history image (MHI) approach is a view-based temporal template method which is simple but robust in representing movements and is widely employed by various research groups for action recognition, motion analysis and other related applications. In this paper, we provide an overview of MHI-based human motion recognition techniques and applications. Since the inception of the MHI template for motion representation, various approaches have been adopted to improve this basic MHI technique. We present all important variants of the MHI method. This paper points some areas for further research based on the MHI method and its variants.

Motion history image: its variants and applications

This paper proposes a superpixel-based spatiotemporal saliency model for saliency detection in videos. Based on the superpixel representation of video frames, motion histograms and color histograms are extracted at the superpixel level as local features and frame level as global features. Then, superpixel-level temporal saliency is measured by integrating motion distinctiveness of superpixels with a scheme of temporal saliency prediction and adjustment, and superpixel-level spatial saliency is measured by evaluating global contrast and spatial sparsity of superpixels. Finally, a pixel-level saliency derivation method is used to generate pixel-level temporal and spatial saliency maps, and an adaptive fusion method is exploited to integrate them into the spatiotemporal saliency map. Experimental results on two public datasets demonstrate that the proposed model outperforms six state-of-the-art spatiotemporal saliency models in terms of both saliency detection and human fixation prediction.

Superpixel-Based Spatiotemporal Saliency Detection

For the last decades, computer-based visual attention models aiming at automatically predicting human gaze on images or videos have exponentially increased. Even if several families of methods have been proposed and a lot of words like centre-surround difference, contrast, rarity, novelty, redundancy, irregularity, surprise or compressibility have been used to define those models, they are all based on the same and unique idea of information innovation in a given context . In this paper, we propose a novel saliency prediction model, called RARE2012, which selects information worthy of attention based on multi-scale spatial rarity. RARE2012 is then evaluated using two complementary metrics, the Normalized Scanpath Saliency (NSS) and the Area Under the Receiver Operating Characteristic (AUROC) against 13 recently published saliency models. It is shown to be the best for NSS metric and second best for AUROC metric on three publicly available datasets (Toronto, Koostra and Jian Li). Finally, based on an additional comparative statistical analysis and the effect-size Hedge' g ⁎ measure, RARE2012 outperforms, at least slightly, the other models while considering both metrics on the three databases as a whole.

RARE2012: A multi-scale rarity-based saliency detection with its comparative statistical analysis

Fast communication: Active energy image plus 2DLPP for gait recognition

Fast communication: Gait recognition based on dynamic region analysis

By the guidance of attention, human visual system is able to locate objects of interest in complex scene. We propose a new visual saliency detection model for both image and video. Inspired by biological vision, saliency is defined locally. Lossy compression is adopted, where the saliency of a location is measured by the Incremental Coding Length(ICL). The ICL is computed by presenting the center patch as the sparsest linear representation of its surroundings. The final saliency map is generated by accumulating the coding length. The model is tested on both images and videos. The results indicate a reliable and robust saliency of our method.

/pdf/incremental-sparse-saliency-detection-1syjgoci5i.pdf

Incremental sparse saliency detection

This invention discloses a method for estimating man-face gestures by a dimensionality reduction method including the following steps: 1, preprocessing man-face image trained samples of different gestures, 2, carrying out PCA process to processed data, 3, initializing limited nerve network of a Boltzmann machine, 4, pre-training the limited Boltzmann machine nerve network with the data processed by PCA, 5, adjusting the limited Boltzmqnn nerve network parameters, 6, identifying gestures of new man-face images, which reduces the error rate even further compared with the present technology and reduces its dimensions greatly.

Method for making human face posture estimation utilizing dimension reduction method

This paper presents a novel method for identity recognition based on the 2D gait representation: Gait Energy Image (GEI) which is the averaged silhouette over one gait cycle. An ensemble of Gabor kernels is first convolved with GEI to extract discriminative feature. The obtained Gabor gait representation is then projected into lower dimensional subspace using discriminative common vectors (DCV) analysis. The final classification is performed in this subspace. The proposed method is tested on the USF HumanID Database. Experimental results show that Gabor-based method can improve recognition rate, and DCV is superior to other traditional dimensional reduction algorithm in the gait recognition application.

Gabor-Based Discriminative Common Vectors for Gait Recognition

This paper proposes an effective gait recognition approach based on the Gait Energy Image (GEI) representation. Synthetic GEI samples are first generated to address the problem of lacking training data. Then the Gabor phase spectrum of GEI which was ignored in the previous work is utilized as the input feature, and it is subsequently embedded into a low dimensional manifold by locality preserving projections to perform classification. The proposed recognition approach is evaluated on the USF HumanID database. Experimental results show that our approach outperforms other automatic algorithms in terms of recognition accuracy.

Xiaochao Yang

Papers

Fast communication: Gait recognition based on dynamic region analysis

Incremental sparse saliency detection

Method for making human face posture estimation utilizing dimension reduction method

Gabor-Based Discriminative Common Vectors for Gait Recognition

Gabor phase embedding of Gait Energy Image for identity recognition