Qubo You

Rotated haar-like features for face detection with in-plane rotation

This paper extends the upright face detection framework proposed by Viola et al. 2001 to handle in-plane rotated faces. These haar-like features work inefficiently on rotated faces, so this paper proposes a new set of ±26.565 ° haar-like features which can be calculated quickly to represent the features of rotated faces. Unlike previous face detection techniques in training quantities of samples to build different rotated detectors, with these new features, we address to build different rotated detectors by rotating an upright face detector directly so as to achieve in-plane rotated face detection. This approach is selected because of its computational efficiency, simplicity and training time saving. This proposed method is tested on CMU-MIT rotated test data and yields good results in accuracy and maintains speed advantage.

50 Years of Image Processing and Pattern Recognition in China

This article briefly reviews the development of image recognition in and outside China. It presents theoretical research achievements and applied research as well as several typical applications of image recognition in China. Finally, it discusses future trends in image recognition integrated with cognitive science. This article is part of a special issue on AI in China.

Facial Expression Sequence Synthesis Based on Shape and Texture Fusion Model

Shape and texture are two aspects in facial expression synthesis. For the requirements of synthesized shape and texture are different, it is unsuitable to use same method to synthesize shape and texture. In this paper, we propose a Statistical Shape and Radio texture fusion model for facial expression sequence synthesis. In this framework, facial shape and texture had been processed separately, then they had been fused together to synthesize final result. The main contributes of this paper are: First propose statistical shape and radio texture fusion model, process shape and texture separately. Second introduce ASM model into facial shape synthesis, and construct expression model of ASM parameters. Third put forward a shape based sample select mechanism, fusion shape and texture process together. Experiment results demonstrate that the proposed model is more expressive and realistic than traditional methods, and can process batch input well.

Eye Synthesis Using the Eye Curve Model

Eyes are a critical part of facial expressions. Because of the appearance diversity of eyes due to motion, it is difficult to synthesize eye with a particular facial expression. Traditional methods have failed to adequately catch motion-related appearance changes. In order to generate a photorealistic expression eye, we propose a two-step method. Firstly, we propose an eye curve model to represent the eye. The model uses one circle and four skewed elliptical arcs to represent the shape of eyes, and divides the entire eye region into 6 sub-regions that correspond to different anatomical components of the eye. Then we propose a structure-based similarity (SBS) framework to synthesize the expression eye using the eye curve model. This paper primarily contributes three things: first, the proposed eye curve model can represent the diversity of eyes, which is better than some traditional models. Second, when there are many samples, our method can synthesize expression eyes with personal style, which is more reasonable when synthesizing common expressions such as joy. Third, when there is only one sample, our method can clone this expression, which is more useful when synthesizing very special expressions such as a grimace. Experimental results show that all synthesized eyes are realistic and expressive.

ANALYSIS OF SOLUTION FOR SUPERVISED GRAPH EMBEDDING

Recently, Graph Embedding Framework has been proposed for feature extraction. However, it is still an open issue on how to compute robust discriminant transformation for this purpose. In this paper, we show that supervised graph embedding algorithms share a general criterion. Based on the analysis of this criterion, we propose a general solution, called General Solution for Supervised Graph Embedding (GSSGE), for extracting the robust discriminant transformation of Supervised Graph Embedding. Then, we analyze the superiority of our algorithm over traditional algorithms. Extensive experiments on both artificial and real-world data are performed to demonstrate the effectiveness and robustness of our proposed GSSGE.

Object Recognition by Learning Informative, Biologically Inspired Visual Features

This paper presents a novel, effective way to improve the object recognition performance of a biologically-motivated model by learning informative visual features. The original model has an obvious bottleneck when learning features. Therefore, we propose a circumspect algorithm to solve this problem. First, a novel information factor was designed to find the most informative feature for each image, and then complementary features were selected based on additional information. Finally, an intra-class clustering strategy was used to select the most typical features for each category. By integrating two other improvements, our algorithm performs better than any other system so far based on the same model.

Solution for supervised graph embedding: A case study

Recently, Graph Embedding Framework has been proposed for feature extraction. Although many algorithms can be used to extract the discriminant transformation of supervised graph embedding, it is still an open issue which algorithm is more robust. In this paper, we first review the classical algorithms which can extract the discriminant transformation of linear discriminant analysis (LDA), and then generalize these classical algorithms for computing the discriminant transformation of supervised graph embedding. Secondly, we theoretically analyze the robustness of these generalized algorithms. Finally, in order to overcome the disadvantages of these generalized algorithms, we propose an effective method, called total space solution for supervised graph embedding (TSS/SGE), to extract the robust discriminant transformation of Supervised Graph Embedding. Extensive experiments and comprehensive comparison on real-world data are performed to demonstrate the robustness of our proposed TSS/SGE.

General Solution for Supervised Graph Embedding

Recently, Graph Embedding Framework has been proposed for feature extraction. However, it is an open issue that how to compute the robust discriminant transformation. In this paper, we first show that supervised graph embedding algorithms share a general criterion (Generalized Rayleigh Quotient). Through novel perspective to Generalized Rayleigh Quotient, we propose a general solution, called General Solution for Supervised Graph Embedding (GSSGE), for extracting the robust discriminant transformation of Supervised Graph Embedding. Finally, extensive experiments on real-world data are performed to demonstrate the effectiveness and robustness of our proposed GSSGE.