Kenji Ohue

Human face profile recognition by a P-Fourier descriptor

A method of recognizing human face profiles is proposed. Most conventional methods of recognizing human face profiles use fiducial marks, length of the line connecting the marks, angles between fiducial points, and other measures of profile outlines as the components of characteristic vectors. The P-Fourier descriptor developed by Uesaka can describe open curves fairly well, and it has many features of curves in the low-frequency range. We use the P-Fourier descriptor as a characteristic vector of a human face profile. We show that the P-Fourier coefficients in the low-frequency range are useful for human face profile recognition, that is, by using 31 P-Fourier coefficients, we obtained a correct recognition rate of 93.1 % for 130 subjects.

Human face recognition by P-type Fourier descriptor

This paper presents a method of recognizing human face profile. The conventional methods of recognizing human face profile use the computer-derived fiducial marks, lines, angles, and other measures of profile outline as the components of characteristic vector. We use P-type Fourier descriptor as a characteristic vector of human face profile. It is shown that four P-type Fourier coefficients in the low frequency range can identify 65 face profiles, with the accuracy of 100%.

Smearing-desmearing transformation and the application to digital watermark

The technology of a digital watermarking that can embed copyright information in valuable images has come to be paid attention. Therefore, we propose a new digital watermarking method using the Smearing-Desmearing transformation and show the algorithm. The Smearing-Desmearing transformation is a new transformation that watermark information is spread on the whole of an image. In our method, we can easily calculate how much the method degrades the quality of the watermarked image from embedded watermark energy.

A 3-D mesh watermarking with uncomplicated frequency selectivity

This paper describes a correlation-based watermarking technique with the fast Fourier transforms (FFTs) for three-dimensional (3-D) mesh models. For generating a watermark with desirable properties, similar to a pseudonoise (PN) signal, an impulse signal on a two-dimensional (2-D) space is spread through the FFT, multiplications of a complex sinusoid signal, and the inverse FFT. This system easily incorporate a frequency selectivity property, because zero-valued components in the multiplication block prevent the energy of the impulse signal at appropriate frequencies. As a result, the proposed approach requires no additional transform, such as the discrete cosine transform and wavelet transform. Since the amount of information that can be stored in the watermark depends on the size of a spread space, the elimination of the subband transform increases a payload of the watermark in a previous paper. The watermark, i.e., spread impulse signal, is embedded into 3-D data aligned by the principle component analysis (PCA). In the detection procedure, after realigning the watermarked mesh model through the PCA, we map the 3-D data on the 2-D space via block segmentation and averaging operation. The 2-D data are processed by the inverse system, i.e., the FFT, the division of the complex sinusoid signal, and the inverse FFT. From the resulting 2-D signal, we detect the position of the maximum value as a signature.

Image coding based on regular cosine-modulated filter banks

In this paper, a novel design method of regular cosine-modulated filter banks (CMFBs) have been presented for image coding. After introducing a regularity constraint on lattice parameters of a prototype filter in paraunitary (PU) CMFBs, we derive a regularity condition for perfect reconstruction (PR) CMFBs. Finally, we design regular 8-channel 32-length PUCMFB and PRCMFB by a unconstrained optimization of residual lattice parameters, and several simulation results for test images are shown for evaluating the proposed image coder based on the CMFBs with one degree of regularity.

On Two-channel orthogonal cyclic filter banks with half-sample symmetric filters

This paper describes a design method of twochannel orthogonal cyclic filter banks with a linear phase for image coding. In the cyclic filter bank, all arguments are interpreted modulo some integer due to the periodicity of signals. This provides more degrees of freedom in design problems. In this paper, we present a systematic method to design two-channel linear phase orthogonal cyclic filter banks with real-valued coefficients. With the aim of employing half-sample symmetric cyclic filters, we derive a condition of the magnitude response for a cyclic filter bank to satisfy the orthogonality. The orthogonal cyclic filter bank with half-sample symmetric filters, which have a nearly ideal magnitude response at discrete values of the frequency, can be designed via the proposed scheme. As one design example, we show such a linear phase orthogonal cyclic filter bank with the period L = 32. Finally, several simulations that compare the performance of our designed cyclic filter bank, the conventional cyclic filter bank and the noncyclic Daubechies 9/7-tap filter bank in image compression are illustrated to evaluate the performance of the proposed approach for images with highly detailed content.