This paper proposes an effective image enlargement method based on an iterated function system (IFS), which is traditionally used for image coding. The IFS can reconstruct an image of a different size from the coding target image's. Based on this property, some methods enlarging images have been proposed by using the IFS. However, the images enlarged by the traditional methods suffer from block noise and edge discontinuity in the boundary between neighboring range blocks, which are units of the process in the IFS. The reasons for the problems of the traditional methods are that they use non-overlapping range blocks and do not consider edge continuity in the boundary between the neighboring range blocks. Therefore, the proposed method allows selection of overlapping range blocks in order to avoid the block noise. Further, the proposed method introduces a line process, which is used for edge detection, into the enlargement procedure. The edges obtained by using the line process can retain the edge continuity. Therefore, the images enlarged by the proposed method retain the edge continuity. Consequently, an accurate image enlargement can be achieved.

This paper proposes an accurate audio signal classification method using a feature extraction matrix. The proposed method classifies the segments of the audio signal into the following five audio classes: silence, speech, music, speech with music background, and speech with noise background. In this classification, a diagonal matrix, which is called the feature extraction matrix, is utilized in order to extract the effective audio features for the classification. By using this feature extraction matrix, the five audio classes are clearly separated each other in the feature space, and thereby highly precise classification can be attained. Experimental results performed by applying the proposed method to real audio signals are shown to verify its high performance.

This paper presents a novel reconstruction method of missing textures using an error reduction algorithm which is one of phase retrieval methods. The proposed method estimates the Fourier transform magnitude of the missing area from another area whose texture is similar in the obtained image. In order to realize this, a novel approach that monitors the errors caused by the error reduction algorithm is introduced into the selection scheme of the similar texture. Further, the proposed method estimates the phase of the target area by using the error reduction algorithm modified for the texture reconstruction and can restore the missing area accurately. Some experimental results show that the proposed method achieves more accurate restoration than that of the traditional methods.

This paper proposes an audio-cut detection and audio-segment classification method using fuzzy c-means clustering. In the proposed method, the boundaries between two different audio signals, which are called audio-cuts, can be detected by fuzzy c-means clustering. In the fuzzy c-means clustering, the fuzzy number represents the possibility that the audio-cut exists. Therefore, according to the possibility, qualified candidates for audio-cuts can be obtained even if audio effects such as fade-in, fade-out, etc. are included in the audio signal. The audio signal is segmented at the detected audio-cuts, and these segments are classified into the following five classes: silence, music, speech, speech with music background, and speech with noise background. This classification simultaneously deletes wrongly detected audio-cuts. Consequently, we can obtain accurate audio-cuts and classification results.

This paper proposes a novel phase retrieval method using active contour models (snakes) for image reconstruction. The proposed method reconstructs a target image from the magnitude of its Fourier transform and the measured area of the image. In general, the measured area is different from the true area where the target image exists. Thus the snake with an area term, which can extract the shape of the concave target image, is utilized to renew the measured area. By processing this renewal iteratively, the area obtained by the snake converges to the true area and as a result the proposed method can accurately reconstruct a target image even when the measured area is different from the true area. Experimental results show the effectiveness of the proposed method. This method has many applications such as biomedical imaging using X-ray or optical detectors which have lost the phase information.

This paper proposes a new noise-detection method for restoration of images corrupted by impulsive noise. The proposed method consists of two stages. In the first stage, the pixels classified according to a new flag image are processed by different noise detectors. They are realized by using two median filters with different sizes of windows. In the second stage, each pixel once detected as an impulse in the first stage is verified by using a new system. According to the above stages, the proposed method can accurately detect the location of the impulsive noise and be effectively used as a preprocessor for noise reduction filtering. Experiments show that the proposed method can effectively detect impulsive noises in noisy images even when they are very highly corrupted.

This paper presents a novel representation method for contour shape using fractal interpolation functions (FIF). In the traditional idea of the FIF, the scope of its application has been limited to the case where the signal is represented by a single-valued function. Therefore, the traditional FIF cannot be applicable to multiple-valued signals. The proposed method can model a multiple-valued signal with an extended FIF derived by introducing new parameters to the traditional one. Furthermore, the proposed method utilizes the fractal dimension known as a measure of complexity to determine the parameters in the FIF and thereby can model the signal based on its complexity. Experimental results show the validity of the proposed method.

This paper proposes a new image restoration method based on the iterated function system (IFS). The proposed method can restore images contaminated by impulsive noise according to self-similarity represented by the IFS parameters. Since the IFS is usually used for image coding, it has never been applied to image restoration; and it cannot be utilized for image restoration as it is. In order to adapt the IFS for image restoration, this paper reforms the conventional criterion for the computation of the IFS parameters to suit for image restoration, and as preprocessing we apply an /spl epsiv/-filter in which a median filter is embedded to contaminated images prior to computing the IFS parameters. Some simulation results are presented to demonstrate the effectiveness of this method.

Multiple Error Filtered X (MEFX) LMS (Least Mean Squares) based booming noise or engine noise cancellation system has been implemented in the automobile and the reciprocate engine airplane cabin to provide practical solutions to improve the annoying noisy environment. The noises are well known in that they behave as resonance noise of the fuselage. A significant difference in noise level or power spectrum between control points is often observed. For the cabin noise cancellation system, all passenger areas are expected to be quiet. Because the MEFX LMS algorithm minimizes the total squared power of the error signals, degradation of increase in the noise level is observed in certain situations. In this paper, we demonstrate such behavior of the algorithm and propose a method to improve the practical performance of the system.

This paper presents an algorithm that is capable of extracting polygons from complex digital images for real-time video-applications. The algorithm uses a high-accuracy, real-time (HART) line-feature extraction algorithm to extract the line segments from the digital image. Then these line segments are grouped into triangles and quadrilaterals using an efficient scanning method that can be easily extended to n-sided polygons. The algorithm can process both convex and concave polygons. Tests were performed on a variety of 256/spl times/256 grayscale synthetic and natural images and an average frame rate of 58 frames per second was achieved using a 450 MHz Pentium II processor. Experiments show that the algorithm accurately extracts the polygons from complex images.

This paper proposes an Iterated Function System (IFS) based image-restoration method. The proposed method can restore images corrupted by impulsive noise according to self-similarity represented by the IFS. Since the IFS is usually used for image coding, it has never been applied to image restoration; and it cannot be utilized for the image restoration as it is. In order to adapt the IFS for image restoration, this paper reforms the conventional criterion for the computation of the IFS parameters to suit for image restoration, and as preprocessing we apply a median filter to the contaminated images prior to computing the IFS parameters. Some simulation results are presented to demonstrate the effectiveness of this method.

This paper presents a method to improve implementation accuracy of a recently proposed CORDIC ARMA lattice filter. Since the CORDIC ARMA lattice filter algorithm has a problem in its shift sequence, it cannot implement a lattice filter accurately. Therefore, in this paper we apply the shift sequence proposed by Walther without problem to the CORDIC ARMA lattice filter, and then we realize an accurate lattice filter. In addition, we propose a new pipelined architecture for the CORDIC ARMA lattice filter. The proposed lattice filter architecture consists of CORDIC processors implemented in the pipelined fashion; it can achieve a high data processing throughput rate.

We propose a practical method to reduce a number of reference signals for an active noise cancellation (ANC) system and the filter characteristics to generate the reduced number of reference signals, which maintain the original value of the coherence function. This method finds the number of independent noise sources and provides the filter characteristics based on SVD (singular value decomposition) of the power spectrum matrix of the reference signals. Then, we also use the multiple coherence function analysis to select dominant components in the reference signals. The method contributes greatly in reducing the number of reference signals for the ANC system that uses the large number of reference signals. We also discuss the characteristics of the filters that synthesizes the new set of reference signals. An experimental test is also performed to confirm the theory.

A new genetic algorithm (GA) based image segmentation method is proposed for image analysis. This method using a mean square error (MSE) based criterion can segment an image into some regions, while estimating a suitable region representation. The criterion is defined as MSE caused by interpolating each region of an observed image with a parametric model. Since the criterion is expressed with not only the parameters of the model but also shape and location of the regions, the criterion can not be easily minimized by the usual optimization methods, the proposed method minimizes the criterion by a GA. The proposed method also includes a processor to eliminate fragile regions with the Markov random field (MRF) model. Though the thresholds of the existent methods negatively affect image segmentation results; since no thresholds are required in the proposed method, it segments images more accurately than the existent methods.

This paper proposes a fractal interpolation for natural images. Generally, linear interpolation and spline interpolation are used for image interpolation. However, an image interpolated by the above conventional methods loses some high-frequency components of an original image. The loss of components lowers fidelity of the interpolated images. Since the proposed method reduces the loss, an interpolated image generated by the proposed method has higher fidelity than the one generated by the conventional method. The reduction of the loss is realized by using the fractional Brownian motion (FBM) in a process of the interpolation. The proposed method uses a characteristic that the fractal dimension is strongly correlated with a sense of roughness.

This paper proposes a high-accuracy quantization method for IFS parameters in fractal image coding by using a genetic algorithm (GA). The development of IFS-parameter quantization techniques is significant for the image coding because its errors make more serious problems in the iteration procedures than the other quantization topics. Even if the errors are small, high-quality reconstructed images are not necessarily obtained. Therefore, the high-accuracy quantization methods are required for the parameters. The proposed method provides higher quality reconstructed images than a conventional method which merely minimizes the errors.

Fractal image compression using the iterative function system (IFS) is based on the collage theorem proposed by Barnsley et al. (1986). In this conventional method, the errors between the reconstructed image and original image may be greater than the errors between the collage and original image because the collage theorem does not guarantee the former errors to be smaller than the latter errors. This paper proposes an extended collage theorem. An IFS algorithm based on this theorem determines the parameters after iterations of the contraction mappings. An image reconstructed according to the new theorem has higher quality than one based on the existing collage theorem. The reconstructed image can be gotten by fewer iterations than by the conventional IFS.

This paper proposes a method for realization of an ARMAS lattice filter. The ARMAX (autoregressive moving average model with exogenous variable) model identification is significant because the ARMAX model is a standard tool in the control field, and it can be performed by the proposed algorithm. One of the recursive least-square methods for the ARMAX model identification is the ELS (extended least squares). Applied to the ARMAX model identification, the ELS uses o(N/sup 2/) multiplications, where N/spl Delta/=AR order+MA order+X order. When the proposed realization method of the ARMAX lattice filter is used, o(M) multiplications are needed for the ARMAX model identification, where M/spl Delta/=ma.

In this paper, a method for realization of an ARMAX lattice filter is proposed. Since the ARMAX model is a standard tool in control, it is significant to solve ARMAX model identification problems and the proposed algorithm can solve it. One of the recursive least-squares methods for the ARMAX model identification is the ELS (Extended Least Squares). If the ELS is applied to the ARMAX model identification, it uses o(N/sup 2/) multiplications, where N /spl Delta//sub =/ AR order + MA order + X order. By using the proposed realization method, o(N) multiplications are needed for the ARMAX model identification.

The residual correlation of the Hadamard transforms of stationary Markov-1 signals is investigated. It is shown that the decorrelation in the Hadamard domain is not substantially improved by increasing the block size or the number of data points. This is in sharp contrast with the DFT and other sinusoidal transforms known for their asymptotic equivalence to the KLT. The computational simplicity of the Hadamard transform is counterbalanced by its poor decorrelating capability. Comparison of the Hadamard and Fourier transforms in their decorrelating performance is included.

A new discrete cosine transform is developed in such a way that its closeness to the Karhunen–Loeve transform can be established straightforwardly. The transform is characterized by a symmetric matrix and will be referred to as the symmetric cosine transform (SCT). It is shown that a simple window on the data vector makes its covariance matrix more tractable for the SCT. Since the derivation of the SCT is, to some extent, similar to that of the fast KLT proposed by Jain, its advantages over the FKLT are discussed. The new transform is also compared with the conventional DCT proposed by Ahmed et al. The performance of the SCT is better than that of the DCT with respect to computational efficiency, the residual correlation, and the rate-distortion criterion. The SCT is also convenient in hardware implementation as both the forward and reverse operations can be handled by a single apparatus.

A few important properties of the fast Karhunen-Loeve transform (FKLT) proposed by Jain [3] are discussed. The purpose here is to emphasize the role of the boundary conditions upon which the entire theory is based. The effects of the zero boundary conditions are also discussed. It is argued that the transform is not as efficient as it might appear when used in bit rate reduction.

In the above correspondence^{1}Hamidi and Pearl proved three theorems concerning asymptotic characteristics of the Fourier coefficients of stationary random signals. From their formulation of Theorem 17 it might appear as though the variances of all the Fourier coefficients converge to a single constant. It is shown that this is not the case.

This concise paper presents a theoretical evaluation of energy packing of the Hadamard transform which is often used in signal processing. It is shown that energy contained in the lowest1/2^{j}(j: positive integer) of a signal's sequency spectrum can be explicitly evaluated in terms of covariances of the signal.