Fengxiang Ge

Weighted joint sparse representation-based classification method for robust alignment-free face recognition

Abstract. This work proposes a weighted joint sparse representation (WJSR)-based classification method for robust alignment-free face recognition, in which an image is represented by a set of scale-invariant feature transform descriptors. The proposed method considers the correlation and the reliability of the query descriptors. The reliability is measured by the similarity information between the query descriptors and the atoms in the dictionary, which is incorporated into the l0∖l2-norm minimization to seek the optimal WJSR. Compared with the related state-of-art methods, the performance is advanced, as verified by the experiments on the benchmark face databases.

Semi-auto assessment system on building damage caused by landslide disaster with high-resolution satellite and aerial images

In recent years, earthquake and heavy rain have triggered more and more landslides, which have caused serious economic losses. The timely detection of the disaster area and the assessment of the hazard are necessary and primary for disaster mitigation and relief. As high-resolution satellite and aerial images have been widely used in the field of environmental monitoring and disaster management, the damage assessment by processing satellite and aerial images has become a hot spot of research work. The rapid assessment of building damage caused by landslides with high-resolution satellite or aerial images is the focus of this article. In this paper, after analyzing the morphological characteristics of the landslide disaster, we proposed a set of criteria for rating building damage, and designed a semi-automatic evaluation system. The system is applied to the satellite and aerial images processing. The performance of the experiments demonstrated the effectiveness of our system.

Automatic landslide and mudflow detection method via multichannel sparse representation

Landslide and mudflow detection is an important application of aerial images and high resolution remote sensing images, which is crucial for national security and disaster relief. Since the high resolution images are often large in size, it’s necessary to develop an efficient algorithm for landslide and mudflow detection. Based on the theory of sparse representation and, we propose a novel automatic landslide and mudflow detection method in this paper, which combines multi-channel sparse representation and eight neighbor judgment methods. The whole process of the detection is totally automatic. We make the experiment on a high resolution image of ZhouQu district of Gansu province in China on August, 2010 and get a promising result which proved the effective of using sparse representation on landslide and mudflow detection.

Improved weighted prediction parameter estimation for advanced video coding systems

Weighted prediction (WP) is an efficient tool for encoding of video with brightness variations, if WP parameters can be estimated accurately. In this paper, an improved estimation method for WP parameters of advanced video coding system (AVS) is proposed. To find a best matching block in reference for current block to be encoded, a motion search scheme is presented. Each block in a searching window in the reference is scrutinized and the WP parameters for the block pair are estimated by least squared. Then the candidate WP parameters of the block to be encoded is select by the WP parameters with minimum sum of absolute difference (SAD) in the search window. Finally, the estimate of WP for current frame is selected by voting of all the candidate WP parameters of all blocks in the frame. The proposed scheme is implemented on the reference model of AVS, and shows that the proposed method can estimate WP parameters more accurate and save more bit rate compared the the realization in reference model (RM) of AVS.

Action recognition based on multiple key motion history images

Human action recognition in video sequences is an important research topic in computer vision, and motion history image (MHI) is widely taken for recognition due to its simplicity. However, it may be not robust to describe an action by only a single MHI. Therefore, an action recognition scheme by using multiple key MHIs (MKMHIs) is proposed. Firstly, an adaptive method for key MHIs selection is proposed based on entropy of MHIs. Then a new combined feature vector of MHIs by spatial pyramid matching (SPM) is defined to describe spatiotemporal characteristics of actions. Here, the proposed solution is composed of SPM two dimensional entropy (2D-entropy) of MHI and SPM Zernike moment of motion history image edge (MHIE), and our combined feature is lower compared with Local Binary Pattern Histogram(LBP_H). At last, action recognition is performed by SVM and voting. Experimental results show the proposed method based on MKMHIs can improve the action recognition ratio.

Damage assessment framework for landslide disaster based on very high-resolution images

Abstract. It is well known that rapid building damage assessment is necessary for postdisaster emergency relief and recovery. Based on an analysis of very high-resolution remote-sensing images, we propose an automatic building damage assessment framework for rainfall- or earthquake-induced landslide disasters. The framework consists of two parts that implement landslide detection and the damage classification of buildings, respectively. In this framework, an approach based on modified object-based sparse representation classification and morphological processing is used for automatic landslide detection. Moreover, we propose a building damage classification model, which is a classification strategy designed for affected buildings based on the spectral characteristics of the landslide disaster and the morphological characteristics of building damage. The effectiveness of the proposed framework was verified by applying it to remote-sensing images from Wenchuan County, China, in 2008, in the aftermath of an earthquake. It can be useful for decision makers, disaster management agencies, and scientific research organizations.

Mode activity based adaptive fast intra mode decision for HEVC/H.265

In the intra coding of HEVC/H.265, the computational complexity of intra mode decision is very high because up to 35 intra prediction modes are supported. Though a fast intra mode decision based on rough mode decision (RMD) is adopted in reference software HM12.0, only a fixed number of intra modes are selected for rate distortion optimization (RDO). In this paper, an adaptive fast intra mode decision for HEVC/H.265 is proposed. We define mode activity by analyzing the rough costs of all modes, and the computational complexity of mode activity is very low. Based on the analysis of relation between mode activity and rate distortion cost gain, we propose to adaptively reduce the number of candidate intra modes for RDO by utilizing the mode activity, so accelerated intra coding is realized. Compared with the RMD in reference code HM12.0 of HEVC/H.265, the proposed algorithm can reduce encoding time with only negligible loss of compression efficiency.

Regularized single-image super-resolution based on progressive gradient estimation

Gradient domain optimization is widely used in regularized image super-resolution, in which the gradient of high resolution (HR) is estimated for calculating the regularization energy. In this paper, a progressive gradient estimation (PGE) is proposed. In PGE, the gradient of the reconstructed HR image in the previous round of optimization is taken as the estimated gradient in the current round. Then, the estimated image gradient is progressively improved. When the estimated image gradient converges, a high quality HR image can be reconstructed. Experimental results show that the reconstructed HR images by PGE have good qualitative and quantitative performances.

SIFT-type descriptors for sparse-representation-based classification

Sparse representation based classification (SRC) as an efficient method has high recognition rate in many pattern recognition applications. Unfortunately, the original SRC method generally requires rigid alignment in classification. In this paper, the feature-based SRC method is addressed by using the PCA-SIFT and SPP-SIFT descriptors, respectively. The presented methods are not only efficient for alignment-free in face and vehicle recognition, but also robust for the image illumination variation, rescaling and affine transform, when the image processing is moved from pixel-domain into the feature-domain and sparse-domain, i.e. PCA-SIFT and SPP-SIFT descriptors. Experimental results show the presented methods in this paper have higher recognition rate, more robustness. In addition, PCA-SIFT-SRC has lower computational complexity than MKD-SRC and SRC in the above scenarios.

Alignment-free sparse representation based classification method via fast location

This paper aims at optimizing the efficiency of the sparse representation based classification (SRC) method in automatic recognition, which is a common problem with large quantity of sample images. An automated target recognition framework based on SRC method is proposed through fast locating to the region of interest (ROI) and dictionary filtering meanwhile. We solve the alignment problem through the fast locating and get an alignment-free SRC method for different poses of a 3D target. We propose two methods for the fast locating in the paper. The dictionary filtering is done according to the probe image. The proposed method has been operated on car and face databases. Car recognition aiming at multi-pose recognition, a car-model database is set up, and its capturing equipment and environments are introduced. On this database, the performance of the proposed method is assessed and compared with the original SRC method. Then, we have further performed the method on yawl B database for face recognition. Then we conclude that the proposed method improves the efficiency and accuracy of the original SRC method.