Search

This paper considers the recognition of moving images and proposes a new framework in which the local features, the global structure, and motion information are handled comprehensively. The method is applied to the extraction of facial expressions and its effectiveness is demonstrated. In most conventional moving image recognition methods, recognition is performed on the basis of the sequence of recognition results for individual frames. However, the authors believe that the motion information should be more positively utilized in the recognition of the individual frames. When the time-series data are combined into the recognition of a still image, however, the amount of information is tremendous, making the processing complicated. The method proposed in this paper extends the concept of the labeled graph matching method, which has been used for still images, to moving images. The proposed method handles sparse graphs and can prevent an increase in the amount of computation. By dynamically adjusting the features to be handled according to the stage of processing, complex processing in dynamic image recognition can be integrated in a simple and straightforward way. As practical examples, the human head, and parts of the head, are extracted, indicating the effectiveness of the proposed method. © 2004 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 87(10): 35–43, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20106 [ABSTRACT FROM AUTHOR]

In this paper, control of emphasis by the multiscale brightness gradient is introduced for unsharp masking. This provides an improved image emphasis method for realizing effective emphasis for any image with superposed Gaussian noise and low-contrast images. Unsharp masking is a method of obtaining an enhanced image by applying a high-pass filter to blurred images so that the obtained high-pass components are superposed on the original image. However, unsharp masking has the deficiency that the noise is also enhanced due to its principle of operation when noise is superposed on the blurred image. In order to alleviate this problem, this paper proposes a characteristic quantity of image edges (edge information) using the amplitude shift between the scales of the multiscale brightness gradient, and applies it to the control of the emphasis of unsharp masking. The proposed edge information takes a positive value on the image edge and the amplitude is proportional to the smoothness and brightness difference of the edges. Since the proposed edge information takes a positive value at the smoothed edges regardless of the variance of the superposed Gaussian noise and the image contrast, it is possible to reduce the effect of noise in the emphasis result by controlling the image emphasis by discriminating positive or negative values. In the experiment, the proposed edge information is introduced into unsharp masking using a fuzzy rule and into third-order unsharp masking. The effectiveness of the proposed edge information is confirmed by image emphasis. © 2003 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 87(4): 40–54, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.10113 [ABSTRACT FROM AUTHOR]

This paper aims at a multiscale image representation which is rotation- and shift-invariant. The wavelet energy distribution is defined, and an image recovery algorithm is proposed based on the representation. In order to derive the image representation, which is almost rotation-invariant, two or three components in different directions of the discrete binary wavelet transform are integrated into a wavelet energy distribution. The image is recovered from the wavelet energy distribution by iterative operations. It is guaranteed in the proposed iterative operation that the error between the wavelet energy distribution of the recovered image and the given wavelet energy distribution decreases and converges. This paper proposes a coarse-to-fine recovery method which is based on the multiscale property of the wavelet transform, and reduces the danger that the procedure is trapped by a local minimum. The usefulness of the proposed coarse-to-fine recovery method is verified through several image recovery examples. As an application example, the method is applied to contour extraction coding by sampling the wavelet energy distribution along the contour of an image. © 1999 Scripta Technica, Electron Comm Jpn Pt 3, 82(12): 65–78, 1999 [ABSTRACT FROM AUTHOR]

Image coding based on the wavelet transform with a lifting configuration has received considerable attention in recent years as a method for unified lossless and nonlossless coding, because effective lossless coding becomes possible by rounding after filter processing. However, if rounding is also performed for lossy coding, rounding errors in addition to quantization errors are superposed on the reconstructed image, so that a further reduction of the SNR takes place. In this paper, a conversion module of a new nonseparable two-dimensional wavelet that can be extended to a general N-stage problem is proposed. In this way, unified lossless and lossy coding with reduced rounding error is possible in addition to interchangeability with the conventional separable wavelet. Further, the propagation mechanism of the rounding error within the filter bank is analyzed. The variance of the rounding errors superposed on the reconstructed image is theoretically derived. In this way, the reduction effect of the rounding error in the N stages is theoretically explained. As a result, the variance of the rounding error is reduced by about 50% in comparison with that for the separable case and an SNR improvement of about 1.30 dB on the average at the time of high-bit-rate lossy coding is confirmed. © 2005 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 88(5): 37–49, 2005; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20124 [ABSTRACT FROM AUTHOR]

The authors propose a method to reproduce surface textures by mapping images used in shape restoration as textures when generating a three-dimensional computer graphic image after restoring a three-dimensional shape from multi-viewpoint data of a real-world model. In the past, an image used as a texture had to be captured under the same lighting conditions and at the same spatial resolution as its source. In this paper, the authors consider a method to select and correct an optimal image from among images captured using free camerawork, and then use it for mapping. In other words, the authors take into consideration the distance between the patch surface and points of capture, while at the same time comparing the orientation of the patch surfaces which comprise a three-dimensional object and the orientation of the camera at the point from which it takes pictures. As a result, optimal pictures can be selected even when the distance to the subject or the position of the camera changes. In addition, by matching the pixel values distributed along the shared boundaries among adjacent patch surfaces, errors that may exist in the mapping between the patch surface and texture image can be corrected. Using their method, the authors perform verification experiments on a three-dimensional shape obtained using the factorization method and achieve good results. © 2004 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 87(5): 70–81, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.10173 [ABSTRACT FROM AUTHOR]

The authors describe a technique for selectively sharpening only edges that were blurred due to various causes without making interference factors such as random noise more conspicuous. The selective sharpening technique described in this paper is implemented as an iterative updating nonlinear algorithm obtained by discretizing a type of nonlinear reaction-diffusion time-evolution equations. The update terms of this algorithm consist of a quadratic nonlinear smoothing term, reaction term, and overshooting term. By performing quantitative evaluation experiments, the authors showed that the proposed technique has significantly better selective sharpness enhancement capabilities than existing sharpness enhancement techniques such as the peaking method. In addition, the authors introduced a processing parameter adaptive control method in the proposed technique and applied it to the suppression of breathing distortions in video sequences. Breathing distortions, which are video sequence distortions that often appear in old films, are time-varying blurring distortions that occur due to temporal variations of the focus. © 2004 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 87(5): 33–47, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.10086 [ABSTRACT FROM AUTHOR]

This paper proposes a method in which the amplitude and phase information of a two-dimensional Gabor filter output are utilized in the supervised segmentation of texture images. The segmentation procedure is as follows. The amplitude and the phase difference between adjacent pixels of the multiple filter outputs are used as the feature parameters. By applying maximum-likelihood estimation of each pixel in the test image, it is decided to which supervised image the pixel belongs. Then, based on the result of segmentation, maximum a posteriori probability estimation is iteratively applied and the final result of segmentation with region integration is obtained. It is shown that when the number of filters in the frequency space is smaller, the phase difference feature parameter contains information which cannot be obtained by using only the amplitude feature parameter. In particular, it is experimentally shown to be effective to weight the phase difference feature parameters by the amplitude. The accuracy of segmentation can be improved by thus combining the phase difference with the amplitude feature parameter. © 2003 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 87(4): 66–79, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.10170 [ABSTRACT FROM AUTHOR]

A new motion-detection model characterized by cooperation between its accommodation and motion-detection mechanisms has been shown to be effective not only in detecting multiple motion velocities but also in theoretically reproducing perceptual responses to moving images. However, the validity of the model's cooperative operation is not necessarily demonstrated. This paper compares the model's motion-detection characteristics with the actual ones including accommodation characteristics for multiple drifting sine waves to verify the cooperative operation. The experimental results support the model's underlying idea of a cooperation between the accommodation and motion-detection mechanisms in human vision. Moreover, the model's cooperation characteristics agree well with the actual ones in the following three aspects: (1) different accommodative states were selected depending on which component wave's motion was the center of attention; (2) at each accommodative state, the attention-getting wave's motion was detected correctly and stably; (3) the degree of the retinal image's blur increased as the attention-getting wave's spatial frequency decreased. These results suggest that the cooperative motion-detection model is valid as a model of the motion-detection mechanism in human vision including the accommodation mechanism. © 2002 Scripta Technica, Electron Comm Jpn Pt 3, 85(5): 35–45, 2002 [ABSTRACT FROM AUTHOR]

This paper proposes a new method of transforming facial expression based on facial sketch image and multi-scale edges. Multiscale edges, which are defined as local maxima of high-frequency signal of facial images, are generally located on the edges of facial parts. Facial sketch images defined as line drawings of edges of facial parts are extracted by morphological processing, the generalized symmetry operator, the rectangle filter, and a geometrical template. Therefore, it is possible to transform the binary sketch image into gray-tone multiscale edge images which are called characterized sketch images. The expression of the characterized facial image can be transformed easily based on FACS (Facial Action Coding System). By reconstruction based on Mallat's method of reconstruction of multiscale edges, facial images with different expressions can be obtained easily. © 2001 Scripta Technica, Electron Comm Jpn Pt 3, 84(9): 67–75, 2001 [ABSTRACT FROM AUTHOR]

In this paper, a method is described for updating the background image corresponding to the environment changing the illumination conditions significantly and for detecting obstructions and tracking of moving vehicles on the road. In the proposed method, the varying region in the monitoring image is derived from the background difference and time difference and is classified into (a) moving objects, (b) stationary objects, and (c) the change due to illumination variation. Based on this classification, region (c) and the region without any change are used for background updating. Also, by forming the a cumulative image in which the binary images composed of regions (a) and (b) are stored for a specified length of time, the detection of the stationary objects is performed and the traffic conditions are determined. For tracking of moving objects, first the DCT is used so that the images consisting only of the moving object and excluding shadow are extracted. Next, these images are used as the template and are tracked by matching. The present method was applied to three types of video images. It was found that all of the 10 stationary vehicles contained in the image were detected. Also, a high accuracy of 93% was accomplished for the tracking of moving objects in the case of daytime images. © 1999 Scripta Technica, Electron Comm Jpn Pt 3, 82(11): 28–37, 1999 [ABSTRACT FROM AUTHOR]

The enlargement of a digital image is a process of narrowing the sampling interval. Since the image before the enlargement and that after the enlargement have different Nyquist frequencies, it is necessary to predict or estimate the high-frequency components that are lost in the image before processing and to complement these components in the enlargement process. It is noted that the image can be represented as the sum of Gaussian components and Laplacian components in a hierarchical structure (pyramid representation). There exists a strong correlation between the Laplacian images of the layers. Based on that property, Greenspan and Anderson have proposed a nonlinear procedure where the unknown Laplacian image of high resolution (i.e., containing the lost high-frequency components) is estimated from the Laplacian image of low resolution. This paper notes that the Greenspan procedure can be network-structured and proposes an enlargement method for digital images using a neural network (NN) for improving the resolution, based on generalizing the network. It is shown that the enlarged NN can be constructed almost independently of the kind of training image or of the hierarchy (resolution). It is also shown that a better enlarged image can be derived compared with Greenspan's method. In other words, the effectiveness and practical usefulness of the proposed method are demonstrated. © 1998 Scripta Technica, Electron Comm Jpn Pt 3, 82(1): 19–29, 1999 [ABSTRACT FROM AUTHOR]

The impression of the human face is very important in daily life. However few studies of it have been made from the technical point of view. This is because impressions of a face are formed in the mind of a person and it is not easy to handle this computationally. This paper proposes a method to synthesize an average face of a group of individuals in order to analyze the “impression” quantitatively. A facial space can be constructed by using multiple average faces. This facial space assists in discriminating among several impressions in changing the impression of the original faces. © 1998 Scripta Technica, Electron Comm Jpn Pt 3, 81(11): 29–35, 1998 [ABSTRACT FROM AUTHOR]

The authors have previously proposed a fuzzy stack filter that is defined by a fuzzy Boolean function. We have also presented a specific design method (called the fuzzy Boolean function determination method) for the most fundamental fuzzy stack filter, the FCWM (fuzzy center weighted median). By varying the fuzzy Boolean function, we can achieve both an arbitrary CWM filter and a weighted average value filter in which only the weights at arbitrary processing points are varied. Hence, in comparison with the CWM filter, our filter is better at eliminating Gaussian noise. In this paper, to improve the reconstruction accuracy of nonsteady state signals, such as the image signals, a data-dependent FCWM (DD-FCWM) is proposed anew. The data-dependent filter is an adaptive filter in which the filter parameters are varied in response to the nature of the data near the processing point (local information). In a DD-FCWM filter, the fuzzy Boolean function is varied by using local information. We demonstrate that the proposed DD-FCWM filter uses the local signal information to appropriately vary, not only the weights of the processing points, but also the filter mode (average value type or median type). Furthermore, by use of image processing examples, we demonstrate that highly accurate signal reconstruction can be attained. © 1998 Scripta Technica, Electron Comm Jpn Pt 3, 81(8): 27–40, 1998 [ABSTRACT FROM AUTHOR]

In the multiple sub-Nyquist sampling encoding (MUSE) system, stationary areas and moving areas of an image are processed differently from each other. These two processings are switched in accordance with the motion detection pixel by pixel. Since this motion detection is carried out independently on the encoder and the decoder, motion detection disagreement may occur between them [“motion detection disagreement”] due to transmission channel noise or particular image patterns. In this paper, the degradation of picture quality caused by the forementioned motion detection disagreement is analyzed. The analysis has made it clear that decoding an image which was encoded as a static image would generate picture degradation accompanied by flicker. However, when a static input image which was encoded as a moving image is decoded as a static image, the decoded image would only generate resolution degradation in its stationary areas but it would not produce any picture degradation through aliasing. This analysis conforms with the results of tests using the MUSE encoder and decoder. Thus, a prototype of a simplified encoder equipped with only a moving image processing was made. This prototype was combined with a MUSE decoder and images of good quality were obtained. The simplified encoder resolution in the stationary areas is somewhat inferior to that of conventional encoders. However, since it helps to reduce the machinery size by a third, or even smaller than that of conventional MUSE encoder, it may be used in consumer Hi-Vision cameras if most of its devices are replaced with LSIs. [ABSTRACT FROM AUTHOR]

One of the highly efficient coding methods for motion pictures is to combine the motion-compensation interframe prediction and the vector quantization. Among such methods, the normalized vector quantization is to represent the motion-compensated interframe difference signal, to separate the average value and to normalize the amplitude. It has a merit in that the general codebook can be utilized. This paper shows that the approximation error does not always decrease in those conventional methods, even if the vector quantization procedure is iterated. As a solution to this problem, it is shown to be sufficient to employ the vector quantizer with a gain/shape structure. An elaboration is made to the codebook size to handle the input vector amplitude. Then the motion picture coding algorithm is constructed using the proposed adaptive gain/shape vector quantizer. By optimizing the coding parameters aiming at the low bit-rate coded transmission, practical temporal and spatial resolutions are obtained for TV conferences and TV phones. [ABSTRACT FROM AUTHOR]

This paper proposes the adaptive discrete cosine transform coding with vector quantization (adaptive DCT-VQ), as a highly efficient coding for the picture signal. In the proposed method, the vector quantization is performed by applying the discrete cosine transform to the picture subblock and partitioning the transformed domain into several vectors. By regarding the transformed domain as more than one vector, the complexity in the coding is reduced. Each vector quantizer is designed according to the standard probability distribution of the cosine transform coefficient, and the universality of the vector quantizer is realized. The mode-selecting type adaptive control is introduced, to effect the adaptive control of the coding parameters according to the local property of the picture. To cope with the large bit allocation resulting from the adaptive control, each vector quantizer has two-stage structure thereby reducing the coding complexity. As a result of the performance evaluation by simulation, it was verified that the proposed coding method has a higher performance than the conventional adaptive cosine transform coding and has much less complexity than the vector quantization in the original space domain. [ABSTRACT FROM AUTHOR]

When interpreting three-dimensional images it is difficult to understand the local concentration of line patterns and hence quantification of the latter is important. In this paper, a concentration index, previously reported to provide an effective characteristic for quantitatively evaluating the local concentration of two-dimensional line patterns, is extended to three dimensions (3D concentration index). The nature and application of this index are described. First, using analysis and theoretical simulation of discrete line patterns, the nature of the 3D concentration index is studied in detail. Then, the 3D concentration index is applied to quantification of the dynamic state of a blood vessel image obtained from a chest X-ray CT image. It is shown that this quantity is effective in the judgment of flux near tumor tissue, a characteristic of a tumor inside the lung. © 1998 Scripta Technica. Electron Comm Jpn Pt 3, 81(7): 7–17, 1998 [ABSTRACT FROM AUTHOR]

This paper describes a novel design of a linear-phase FIR filter and its hardware. The proposed filter has a hierarchical building block structure of tapped-cascaded lower-order subfilters. Since this architecture can reduce the sensitivity of the rounded filter coefficients, an effective high-speed processing of the subfilter can be performed by using adders instead of multipliers. Also, identical subfilters are arranged regularly so that the architecture is suitable for LSI implementation. Another important feature is that the design of the higher-order FIR filter can be replaced with that of the lower-order subfilter. An LSI implementation of about a 45-order FIR filter also is described. [ABSTRACT FROM AUTHOR]

Formant estimation from a speech signal that is observed acoustically is import- and for speech analysis and synthesis. This paper proposes an adaptive formant estimation method with compensation for the gross spectral shape. This estimator can estimate the center frequencies and the bandwidths of formants (anti-formants) based directly on the pole-zero transfer function represented by the products of the second- and first-order factors. Since the proposed method can compensate the gross spectral shape adaptively, excellent formant estimation is expected for continuous speech whose gross spectral shape is time-varying. Therefore, the method proposed herein is applied to the synthetic signal and real speech, and the validity of this method is demonstrated. Furthermore, the Single-Stage Iteration Filter is the estimation algorithm introduced for use in the proposed method. It is shown that the estimation results of tracking with little degradation are obtained stably in the transient part of the spectra. [ABSTRACT FROM AUTHOR]

This paper offers a novel fuzzy filter structure for edge-preserving smoothing of an image corrupted by impulsive and white Gaussian noise. This filter structure is expressed as an adaptive weighted mean filter that uses fuzzy control. The coefficients of the proposed filter can be varied adaptively by fuzzy control laws based on differences between pixels in the window. In the setting of fuzzy control laws, the method takes the distance between the center and each pixel in the window into account. Finally, simulation results demonstrate the effectiveness of the proposed technique. © 1998 Scripta Technica, Electron Comm Jpn Pt 2, 81(1): 47–55, 1998 [ABSTRACT FROM AUTHOR]