Wen-Hsing Hsu

Fast algorithm for point pattern matching : Invariant to translations, rotations and scale changes

Abstract Based on 2-D cluster approach, a fast algorithm for point pattern matching is proposed to effectively solve the problems of optimal matches between two point pattern under geometrical transformation and correctly identify the missing or spurious points of patterns. Theorems and algorithms are developed to determine the matching pairs support of each point pair and its transformation parameters (scaling s and rotation ϑ) on a two-parameter space ( s,ϑ ). Experiments are conducted both on real and synthetic data. The experimental results show that the proposed matching algorithm can handle translation, rotation, and scaling differences under noisy or distorted condition. The computational time is just about 0.5 s for 50 to 50 point matching on Sun-4 workstation.

A modified fast parallel algorithm for thinning digital patterns

Abstract A modified version of the fast parallel thinning algorithm proposed by Zhang and Suen is presented in this paper. It preserves the original merits such as the contour noise immunity and good effect in thinning crossing lines; and overcomes the original demerits such as the serious shrinking and line connectivity problems.

A Video Watermarking Technique Based on Pseudo-3-D DCT and Quantization Index Modulation

The increasing popularity of the internet means that digital multimedia are transmitted more rapidly and easily. And people are very aware for media ownership. However, digital watermarking is an efficient and promising means to protect intellectual properties. Based on the intellectual property attention in the information era, how to protect the personal ownership is extremely important and a necessary scheme. In this paper, we propose an effective video watermarking method based on a pseudo-3-D discrete cosine transform (DCT) and quantization index modulation (QIM) against several attacks. The watermark is mainly inserted into the uncompressed domain by adjusting the correlation between DCT coefficients of the selected blocks, and the watermark extraction is blind. This approach consists of a pseudo-3-D DCT, watermark embedding, and extraction. A pseudo-3-D DCT, which is taken DCT transformation twice, will be first utilized to calculate the embedding factor and to obtain the useful messages. Using the QIM, we embed the watermark into the quantization regions from the successive raw frames in the uncompressed domain and record the relative information to create a secret embedding key. This secret embedding key will further apply to extraction. Experimental results demonstrate that the proposed method can survive filtering, compressions, luminance change, and noise attacks with a good invisibility and robustness.

Accurate optical flow computation under non-uniform brightness variations

In this paper, we present a very accurate algorithm for computing optical flow with non-uniform brightness variations. The proposed algorithm is based on a generalized dynamic image model (GDIM) in conjunction with a regularization framework to cope with the problem of non-uniform brightness variations. To alleviate flow constraint errors due to image aliasing and noise, we employ a reweighted least-squares method to suppress unreliable flow constraints, thus leading to robust estimation of optical flow. In addition, a dynamic smoothness adjustment scheme is proposed to efficiently suppress the smoothness constraint in the vicinity of the motion and brightness variation discontinuities, thereby preserving motion boundaries. We also employ a constraint refinement scheme, which aims at reducing the approximation errors in the first-order differential flow equation, to refine the optical flow estimation especially for large image motions. To efficiently minimize the resulting energy function for optical flow computation, we utilize an incomplete Cholesky preconditioned conjugate gradient algorithm to solve the large linear system. Experimental results on some synthetic and real image sequences show that the proposed algorithm compares favorably to most existing techniques reported in literature in terms of accuracy in optical flow computation with 100% density.

Adaptive thresholding algorithm and its hardware implementation

Abstract This paper presents an adaptive raster-scan thresholding algorithm which can deal with an image acquired under imperfect illumination. A statistical measurement called LSSD (Largest Static State Difference) relating to the gray-level variation is found first. According to the measurement, the pixels are separated into static and transient states which are treated by two different procedures, respectively. A hardware implementation of this algorithm shows that the real-time requirement can be met. Experiments of applying this algorithm to extracting characters from documents confirmed that a reasonable binary image can be efficiently and effectively obtained from a gray-level image under various illuminations.

A systematic approach for designing 2-subcycle and pseudo 1-subcycle parallel thinning algorithms

Abstract This paper describes a systematic approach for designing parallel thinning algorithms, in which three new functions, named local connecting, extended local connecting and erosive direction number, are introduced. With these functions as well as two properties of shape invariance of local edges and local straight lines, all the possible cases of 2-subcycle parallel thinning algorithm are constructed and all the corresponding removing conditions are generated and assigned automatically. In addition, the pseudo 1-subcycle parallel thinning algorithm is also presented. Finally, the effects and efficiency of the above proposed algorithms are analyzed and compared with those of some presently well-known algorithms. Experimental results confirm this new approach, and an efficient and effective algorithm has been built for practical applications.

Fuzzy approach to solve the recognition problem of handwritten Chinese characters

Abstract A method based on the concept of fuzzy set for handwritten Chinese character (HCC) recognition is proposed in this paper. Chinese characters can be viewed as a collection of line segments, called strokes. Since the strokes under consideration here are fuzzy in nature, the concept of fuzzy set is utilized in the similarity measure. Two membership functions are defined for the location measure and type measure between two strokes, and a function of fuzzy entropy is used in information measure. Although the recognition problem can be reduced to the assignment problem, some modifications are still necessary. All the similarities between the corresponding strokes can be chosen by solving the assignment problem using the cost function of fuzzy entropy, and then are averaged to derive the score of similarity between two Chinese characters. 881 classes of Chinese characters in ETL-8 (160 variations/class) are used as the test patterns, and the recognition rate is about 96%. In addition, experiments about the effects of the membership function based on the class separability are also discussed in this paper.

Recognition of handprinted Chinese characters via stroke relaxation

Abstract A new relaxation matching method based on the information of the neighborhood relationship among extracted sub-strokes is proposed to recognize handprinted Chinese characters (HCCs). In order to ensure the convergence in the relaxation process, a new iterated scheme is devised. A supporting function is also designed to solve the problem of wide variability among writers and some inevitable defects in the preprocessing procedure. The distance function on which the matching possibilities of sub-strokes are reflected is determined by using the linear programming method to obtain the best result. The experiments are conducted by using the Kanji of the ETL-8 database. From the experimental results, it is shown that the proposed algorithm does improve the recognition rate of HCCs.

Parallel algorithm for corner finding on digital curves

Abstract A parallel algorithm for corner finding on the skeletons of digital curves is described in this paper. Instead of computing the abrupt change of slopes (discontinuity) with the chain code sequence, the degree of bending via an extended ( k -step) 3 × 3 mask is considered in corner finding.

An interpretive model of line continuation in human visual perception

Abstract Continuation is a fundamental intuitive property of perceptual grouping in line segregation. To explain this property, a hypothetical model from the standpoint of computer engineering is presented in this paper. This model includes four mechanisms (i.e. signal beaming mechanism, feedback transmitting mechanism, image partitioning mechanism, and image revising mechanism) and a temporary 3-D space (called rho-space or quotient space). The map of excitatory and inhibitory signals from an input image is first constructed by applying some functions of the so-called simple cortical cells which are sensitive to the position and orientation of the stimulus. The signal beaming mechanism and feedback transmitting mechanism interact (in human behaviours such as walking) to resolve problems such as continuity, transition, fuzziness, and ambiguity. By using these two mechanisms, the constructed map is then converted into a three-dimensional quotient space in which all the members (voxels) satisfy the equivalence relationships so that the image partitioning mechanism can accomplish the work of line segregation. The image revising mechanism is further used to eliminate the “jaggedness” caused by the signal transmission so that the final result of line segregation is consistent with the one from human observation. This interpretive model was built on a 32-bits workstation (SUN-3/160). The results of several image processing simulations suggest that the hypothetical model is feasible and reasonable.