Thomas C. Chuang

Palmprint verification using hierarchical decomposition

A reliable and robust personal verification approach using palmprint features is presented in this paper. The characteristics of the proposed approach are that no prior knowledge about the objects is necessary and the parameters can be set automatically. In our work, a flatbed scanner is adopted as an input device for capturing palmprint images; it has the advantages of working without palm inking or a docking device. In the proposed approach, two finger-webs are automatically selected as the datum points to define the region of interest (ROI) in the palmprint images. The hierarchical decomposition mechanism is applied to extract principal palmprint features inside the ROI, which includes directional and multi-resolution decompositions. The former extracts principal palmprint features from each ROI. The latter process the images with principal palmprint feature and extract the dominant points from the images at different resolutions. A total of 4800 palmprint images were collected from 160 persons to verify the validity of the proposed palmprint verification approach and the results are satisfactory with acceptable accuracy (FRR: 0.75% and FAR: 0.69%). Experimental results demonstrate that our proposed approach is feasible and effective in palmprint verification.

A novel approach to the detection of small objects with low contrast

This paper proposes an effective approach to the detection of small objects by employing watershed-based transformation. In our work, moving objects with small size and low contrast are first detected from an image sequence which was captured from a video camera. The proposed detection system includes two main modules, region of interest (ROI) locating and contour extraction. In the former module, an image differencing technique is first employed on two contiguous image frames to generate rough candidate objects appearing in the images. A novel neighboring encoding technique along with the image differencing technique is devised here to effectively reduce noise which usually affects the performance of detection results, especially for small objects. Next, we find the bounded rectangles enclosing the denoised candidate objects, which in turn generate ROI. However, the results of the previous process fail to characterize object contours. To do this, we need to devise a contour extraction technique. Unfortunately, satisfactory results cannot be obtained by applying traditional contour extraction methods. To solve this problem, the watershed-based transformation along with the region matching technique is employed to obtain better object contours. Experimental results validate that the proposed approach is indeed feasible and effective in detecting objects with small size and low contrast.

Designing asymmetric Hopfield-type associative memory with higher order hamming stability

The problem of optimal asymmetric Hopfield-type associative memory (HAM) design based on perceptron-type learning algorithms is considered. It is found that most of the existing methods considered the design problem as either 1) finding optimal hyperplanes according to normal distance from the prototype vectors to the hyperplane surface or 2) obtaining weight matrix W=[w/sub ij/] by solving a constraint optimization problem. In this paper, we show that since the state space of the HAM consists of only bipolar patterns, i.e., V=(v/sub 1/,v/sub 2/,...,v/sub N/)/sup T//spl isin/{-1,+1}/sup N/, the basins of attraction around each prototype (training) vector should be expanded by using Hamming distance measure. For this reason, in this paper, the design problem is considered from a different point of view. Our idea is to systematically increase the size of the training set according to the desired basin of attraction around each prototype vector. We name this concept the higher order Hamming stability and show that conventional minimum-overlap algorithm can be modified to incorporate this concept. Experimental results show that the recall capability as well as the number of spurious memories are all improved by using the proposed method. Moreover, it is well known that setting all self-connections w/sub ii//spl forall/i to zero has the effect of reducing the number of spurious memories in state space. From the experimental results, we find that the basin width around each prototype vector can be enlarged by allowing nonzero diagonal elements on learning of the weight matrix W. If the magnitude of w/sub ii/ is small for all i, then the condition w/sub ii/=0/spl forall/i can be relaxed without seriously affecting the number of spurious memories in the state space. Therefore, the method proposed in this paper can be used to increase the basin width around each prototype vector with the cost of slightly increasing the number of spurious memories in the state space.

TotalRecall: A Bilingual Concordance for Computer Assisted Translation and Language Learning

This paper describes a Web-based English-Chinese concordance system, Total-Recall, developed to promote translation reuse and encourage authentic and idiomatic use in second language writing. We exploited and structured existing high-quality translations from the bilingual Sinorama Magazine to build the concordance of authentic text and translation. Novel approaches were taken to provide high-precision bilingual alignment on the sentence, phrase and word levels. A browser-based user interface (UI) is also developed for ease of access over the Internet. Users can search for word, phrase or expression in English or Chinese. The Web-based user interface facilitates the recording of the user actions to provide data for further research.

Aligning Parallel Bilingual Corpora Statistically with Punctuation Criteria

We present a new approach to aligning sentences in bilingual parallel corpora based on punctuation, especially for English and Chinese. Although the length-based approach produces high accuracy rates of sentence alignment for clean parallel corpora written in two Western languages, such as French-English or German-English, it does not work as well for parallel corpora that are noisy or written in two disparate languages such as Chinese-English. It is possible to use cognates on top of the length-based approach to increase the alignment accuracy. However, cognates do not exist between two disparate languages, which limit the applicability of the cognate-based approach. In this paper, we examine the feasibility of exploiting the statistically ordered matching of punctuation marks in two languages to achieve high accuracy sentence alignment. We have experimented with an implementation of the proposed method on parallel corpora, the Chinese-English Sinorama Magazine Corpus and Scientific American Magazine articles, with satisfactory results. Compared with the length-based method, the proposed method exhibits better precision rates based on our experimental reuslts. Highly promising improvement was observed when both the punctuation-based and length-based methods were adopted within a common statistical framework. We also demonstrate that the method can be applied to other language pairs, such as English-Japanese, with minimal additional effort.

Adaptive Bilingual Sentence Alignment

We present a new approach to the problem of aligning English and Chinese sentences in a bilingual corpus based on adaptive learning. While using length information alone produces surprisingly good results for aligning bilingual French and English sentences with success rates well over 95%, it does not fair as well for the alignment of English and Chinese sentences. The crux of the problem lies in greater variability of lengths and match types of the matched sentences. We propose to cope with such variability via a two-pass scheme under which model parameters can be learned from the data at hand. Experiments show that under the approach bilingual English-Chinese texts can be aligned effectively across diverse domains, genres and translation directions with accuracy rates approaching 99%.

LED Backlight Module by a Lightguide-Diffusive Component With Tetrahedron Reflector Array

In our previous work, IEEE/OSA Journal of Display Technology, February 2012, we examined the luminance and uniformity characteristics of a newly invented secondary optical lens with a wide emission angle called a “lightguide-diffusive component”. The lightguide-diffusive component is designed for thin direct LED backlighting applications. The LED backlight module is composed of at least an LED light source, a secondary lightguide-diffusive component with a micro-structure reflective bottom surface, and a flexible printed circuit (FPC) for LED lighting without a brightness-enhanced film (BEF). Through this lightguide-diffusive component, the emission profile of a single LED is modified to provide better illumination and more uniformity. In this work, the micro-structure reflective bottom is redesigned to be shaped like a Tetrahedron Reflector Structure Array instead of a triangle cylinder structured array, resulting in a more uniform spatial light energy distribution on the emission plane of the LED backlight. The simulation results show an increase in the uniformity ratio with the new design from 72% to 76%, a 4% improvement, the angle of the luminous intensity is increased from 98 deg to 112 deg, a 14 deg improvement, a decrease in the luminance is from 9024 nits to 8551 nits, an impairment of 5%, and a decrease in the volume of the micro-structure reflector from 1.154 mm3 to 0.767 mm3, a 33.5% volume reduction (cost saving).

Data hiding of binary images using pair-wise logical computation mechanism

Due to the emergence of digital libraries and the fast development of multimedia, more and more people use data hiding techniques to hide annotations or side information in images. We propose a novel data hiding mechanism based on pair-wise logical computation. The proposed mechanism can achieve the benefits of reversible and lossless reconstruction of the hidden data and the host image without utilizing any information from the original host image. It does not degrade the visual quality of the recovered host image after extracting the hidden data. Moreover, satisfactory data hiding capacity can be obtained simultaneously. The proposed data hiding mechanism is suitable for applying to the data hiding of images, scanned texts, figures, and signatures, especially for side information and annotation data embedding

Efficient multi-resolution histogram matching for fast image/video retrieval

Most content-based image/video retrieval systems use histogram matching method to compute the similarity between two histograms. The matching of two images can be accomplished by matching their corresponding histograms. A good image/video retrieval system requires two factors: fast response time and high accuracy. A fast search algorithm called MRSA was proposed previously by applying a multi-resolution structure to gain speed-up and to have the same retrieval accuracy as the exhaustive search algorithm. In this paper, we improve the retrieving speed of MRSA while maintaining the global retrieval accuracy. The retrieving speed is improved by using the non-uniform quantization method to obtain lower resolution histograms and the non-uniform quantization method is proven to be able to reduce the number of comparisons at lower resolution levels. Furthermore, we not only extend the multi-resolution concept from uniform quantization to non-uniform quantization but also employ another similarity measurement, @g^2distance, to construct the multi-resolution structure. Due to the thresholding mechanism, the proposed non-uniform quantization based method relieves the over-smooth problem suffering from downsampling. Hence, our method will reduce noticeable unnecessary comparisons at low resolution levels than MRSA by selecting a proper quantization table. The employing of additional similarity measurement and different quantization criterion increases the flexibility and the efficiency of the algorithm. Experiments demonstrate the validity and efficiency of our algorithm in some typical image/video retrieval applications.

Alignment of Bilingual Named Entities in Parallel Corpora Using Statistical Model

Named entities make up a bulk of documents. Extracting named entities is crucial to various applications of natural language processing. Although efforts to identify named entities within monolingual documents are numerous, extracting bilingual named entities has not been investigated extensively owing to the complexity of the task. In this paper, we describe a statistical phrase translation model and a statistical transliteration model. Under the proposed models, a new method is proposed to align bilingual named entities in parallel corpora. Experimental results indicate that a satisfactory precision rate can be achieved. To enhance the performance, we also describe how to improve the proposed method by incorporating approximate matching and person name recognition. Experimental results show that performance is significantly improved with the enhancement.