Xiaozheng Zhang

Bio: Xiaozheng Zhang is an academic researcher from Griffith University. The author has contributed to research in topics: Facial recognition system & Three-dimensional face recognition. The author has an hindex of 9, co-authored 30 publications receiving 812 citations. Previous affiliations of Xiaozheng Zhang include Nanyang Technological University & University of Connecticut.

Recognizing Rotated Faces From Frontal and Side Views: An Approach Toward Effective Use of Mugshot Databases

Abstract: Mug shot photography has been used to identify criminals by the police for more than a century. However, the common scenario of face recognition using frontal and side-view mug shots as gallery remains largely uninvestigated in computerized face recognition across pose. This paper presents a novel appearance-based approach using frontal and sideface images to handle pose variations in face recognition, which has great potential in forensic and security applications involving police mugshot databases. Virtual views in different poses are generated in two steps: 1) shape modelling and 2) texture synthesis. In the shape modelling step, a multilevel variation minimization approach is applied to generate personalized 3-D face shapes. In the texture synthesis step, face surface properties are analyzed and virtual views in arbitrary viewing conditions are rendered, taking diffuse and specular reflections into account. Appearance-based face recognition is performed with the augmentation of synthesized virtual views covering possible viewing angles to recognize probe views in arbitrary conditions. The encouraging experimental results demonstrated that the proposed approach by using frontal and side-view images is a feasible and effective solution to recognizing rotated faces, which can lead to a better and practical use of existing forensic databases in computerized human face-recognition applications.

Removal of threading dislocations from patterned heteroepitaxial semiconductors by glide to sidewalls

Abstract: We have shown that threading dislocations can be removed from patterned heteroepitaxial semiconductors by glide to the sidewalls, which is driven by the presence of image forces. In principle, it should be possible to attain highly mismatched heteroepitaxial semiconductors which are completely free from threading dislocations, even though they are not pseudomorphic, by patterned heteroepitaxial processing. There are two basic approaches to patterned heteroepitaxial processing. The first involves selective area growth on a pre-patterned substrate. The second approach involves post-growth patterning followed by annealing. We have developed a quantitative model which predicts that there is a maximum lateral dimension for complete removal of threading dislocations by patterned heteroepitaxy. According to our model, this maximum lateral dimension is proportional to the layer thickness and increases monotonically with the lattice mismatch. For heteroepitaxial materials with greater than 1% lattice mismatch, our model predicts that practical device-sized threading dislocation-free regions may be realized by patterned heteroepitaxial processing.

Automatic Texture Synthesis for Face Recognition from Single Views

Abstract: One possible solution for pose- and illumination-invariant face recognition is to employ appearance-based approaches, which rely greatly on correct facial textures. However, existing facial texture analysis algorithms are suboptimal, because they usually neglect specular reflections and require numerous training images for virtual view synthesis. This paper presents a novel texture synthesis approach from a single frontal view for face recognition. Using a generic 3D face shape, facial textures are analyzed with consideration of all of the ambient, diffuse, and specular reflections. Virtual views are synthesized under different poses and illuminations. The proposed approach was evaluated using the CMU-PIE face database. Encouraging results show that the proposed approach improves face recognition performances across pose and illumination variations

Complete removal of threading dislocations from mismatched layers by patterned heteroepitaxial processing

Abstract: We report the effectiveness of patterned heteroepitaxial processing (PHP) in removing threading dislocations (TDs) from ZnSe epitaxial layers grown on GaAs substrates by metalorganic vapor phase epitaxy. The PHP approach used here involves postgrowth patterning of continuous epitaxial layers followed by annealing. In this study, each as-grown ZnSe/GaAs sample was first cut into pieces forming four types of samples, namely: (1) as grown, (2) postgrowth annealed, (3) postgrowth patterned, and (4) PHP prepared (patterned and annealed). The epitaxial layers with thicknesses of 2000–6000 A were patterned to create 500–6000-A-high and 3–70-μm-wide square mesas that were separated by 20 μm trenches. TD densities were determined by the etch pit density (EPD) technique and comparisons were made between the four types of samples. The first three types of samples exhibited EPDs of approximately 107 cm−2, which indicate that neither patterning alone nor annealing alone was effective at reducing TDs. In contrast, PHP ...

A reflectance model for computer graphics

Abstract: This paper presents a new reflectance model for rendering computer synthesized images. The model accounts for the relative brightness of different materials and light sources in the same scene. It describes the directional distribution of the reflected light and a color shift that occurs as the reflectance changes with incidence angle. The paper presents a method for obtaining the spectral energy distribution of the light reflected from an object made of a specific real material and discusses a procedure for accurately reproducing the color associated with the spectral energy distribution. The model is applied to the simulation of a metal and a plastic.

Learning to generate chairs with convolutional neural networks

Abstract: We train a generative convolutional neural network which is able to generate images of objects given object type, viewpoint, and color. We train the network in a supervised manner on a dataset of rendered 3D chair models. Our experiments show that the network does not merely learn all images by heart, but rather finds a meaningful representation of a 3D chair model allowing it to assess the similarity of different chairs, interpolate between given viewpoints to generate the missing ones, or invent new chair styles by interpolating between chairs from the training set. We show that the network can be used to find correspondences between different chairs from the dataset, outperforming existing approaches on this task.