This paper presents an overview of research activities in Japan in the field of very low bit-rate video coding. Related research based on the concept of "intelligent image coding" started in the mid-1980's. Although this concept originated from the consideration of a new type of image coding, it can also be applied to other interesting applications such as human interface and psychology. On the other hand, since the beginning of the 1990's, research on the improvement of waveform coding has been actively performed to realize very low bit-rate video coding. Key techniques employed here are improvement of motion compensation and adoption of region segmentation. In addition to the above, we propose new concepts of image coding, which have the potential to open up new aspects of image coding, e.g., ideas of interactive image coding, integrated 3-D visual communication and coding of multimedia information considering mutual relationship amongst various media.

The estimation of model parameter is essentially important for an MRF image model to work well. Because the maximum likelihood estimate (MLE), which is statistically optimal, is too difficult to implement, the conventional estimates such as the maximum pseudo-likelihood estimate (MPLE), the coding method estimate (CME), and the least-squares estimate (LSE) are all based on the (conditional) pixel probabilities for simplicity. However, the conventional pixel-based estimators are not very satisfactorily accurate, especially when the interactions of pixels are strong. We therefore propose two window-based estimators to improve the estimation accuracy: the adjoining-conditional-window (ACW) scheme and the separated-conditional-window (SCW) scheme. The replacement of the pixel probabilities by the joint probabilities of window pixels was inspired by the fact that the pixels in an image present information in a joint way and hence the more pixels we deal with the joint probabilities of, the more accurate the estimate should be. The window-based estimators include the pixel-based ones as special cases. We present respectively the relationship between the MLE and each of the two window-based estimates. Through the relationships we provide a unified view that the conventional pixel-based estimates and our window-based estimates all approximate the MLE. The accuracy of all the estimates can be described by two types of superiority: the cross-scheme superiority that an ACW estimate is more accurate than the SCW estimate with the same window size, and the in-scheme superiority that an ACW (or SCW) estimate more accurate than another ACW (or SCW) estimate which uses smaller window size. The experimental results showed the two types of superiority and particularly the significant improvement in estimation accuracy due to using window probabilities instead of pixel probabilities.

It is shown that five optimal and one quasioptimal binary codes with respect to the Griesmer bound can be obtained from cyclic codes over GF(2fm). An [m(2em - 1), em, 2em-1m] code, a [3(22e - 1), 2e, 322e-1] code, a [2(22e - 1), 2, (22e+2 - 4)/3] code, a [3(22e - 1), 2, 22e+1 - 2] code, and a [3(22e - 1), 2(e+1), 322e-1 - 2] code are optimal and a [2(22e - 1), 2(e + 1), 22e - 2] code is quasi-optimal.

A new type of noncausal stochastic model is proposed to represent a random image with double peak spectrum. The model based on the assumption that the double peak spectrum is expressed by a product of two spectra located at two symmetric positions in the 2D spatial frequency space. Estimation of model parameters is made by means of minimizing the "whiteness" which was proposed in authors' previous work. In a simulation for model estimation we make use of computer-generated random images with double peak spectrum. Comparing this with the estimation by a causal model, we demonstrate that the present method can better estimate not only the spectral peak location but also the spectral shape. The proposed model can be extend to an image model with multl-peak spectrum. However, Increase of parameters makes the model estimation more difficult We try a model with triple peak spectra since a real texture image usually possesses a spectral peak at the origin besides the two peaks. A result shows that the estimation of three spectral positions are good enough, but their spectral shapes are not necessarily satisfactory. It is expected that the estimation of multi-peaked spectral model can be made better by improving the process of minimizing the "whiteness."

Introducing a mathematical model of image noise, we formalize the problem of fitting a conic to point data as statistical estimation. It is shown that the reliability of the fitted conic can be evaluated quantitatively in the form of the covariance tensor. We present a numerical scheme called renormalization for computing an optimal fit and at the same time evaluating its reliability. We also present a scheme for visualizing the reliability of the fit by means of the primary deviation pair. Our method is illustrated by showing simulations and real-image examples.

When an image of a 3D object is transmitted or recorded, its range image as well its grey-scale image are required. In this paper, we propose a method of coding for efficient compression of a pair of a pair of range and grey-scale images of 3D objects. We use Lambertian reflection optical model to model the relationship between a 3D object shape and it's brightness. Good illuminant direction estimation leads to good grey-scale image generation and furthermore effects compression results. A method for estimating the illuminant derection and composite albedo from grey-scale image statistics is presented. We propose an approach for estimating the slant angle of illumination based on an optical model from a pair of range and grey-scale images. Estimation result shows that our approach is better. Using the estimated parameters of illuminant direction and composite albedo a synthetic grey-scale image is generated. For comparison, a comparison coding method is used, in which we assume that the range and grey-scale images are compressed separately. We propose an efficient compression coding method for a pair of range and grey-scale images in which we use the correlation between range and grey-scale images, and compress them together. We also evaluate the coding method on a workstation and show numerical results.

A new application of the factorization method is reported for 3-D shape reconstruction from endoscope image sequences. The feasibility of the method is verified with some theoretical considerations and results of extensive experiments. This method was developed by Tomasi and Kanade, and improved by Poelman and Kanade, with the aim of achieving accurate shape reconstruction by using a large number of points and images, and robustly applying well-understood matrix computations. However, the latter stage of the method, called normalization, is not as easily understandable as the use of singular value decomposition in the first stage. In fact, as shown in this report, many choices are possible for this normalization and a variety of results have been obtained depending on the choice. This method is easy to understand, easy to implement, and provides sufficient accuracy when the approximation used for the optical system is reasonable. However, the details of the theoretical basis require further study.

This paper addresses the problem of estimating 3-D motion of a rigid object from a sequence of monocular 2-D images. The surface of object is assumed to be modeled with several patches, each of which is expressed by an implicit equation. The proposed method estimates the pose (i.e., the location and orientation) of object that corresponds to each image in the sequence: The sequence of the estimated poses gives the motion of the object. The estimation is done by solving a system of equations, each of which is typically an algebraic equation of low degree, that is derived from the expressions of the surface patches and image contours data: so the method does not require establishing the correspondence between successive two frames in the image sequence or computing optic flow. Allowing several-patch models for objects enables the proposed approach to deal with a great variety of objects. The paper includes a numerical example, where our aproach has been applied to a polyhedral object modeled with several patches.

An algorithm for finding the unequal error protection (UEP) capability of a q-ary image of a low-rate qm-ary cyclic code is presented by combining its concatenated structure with the UEP capability of concatenated codes. The results are independent of a choice of a basis to be used for expanding an element over GF (qm) into GF (q). A table of the UEP capability of binary images of low-rate Reed-Solomon codes over GF (26) is given. It is shown that the encoding and decoding algorithms for superimposed concatenated codes can be applied to a q-ary image as a linear UEP code.

Introducing a mathematical model of image noise, we formalize the problem of fitting a line to point data as statistical estimation. It is shown that the reliability of the fitted line can be evaluated quantitatively in the form of the covariance matrix of the parameters. We present a numerical scheme called renormalization for computing an optimal fit and at the same time evaluating its reliability. We also present a scheme for visualizing the reliability of the fit by means of the primary deviation pair and derive an analytical expression for the reliability of a line fitted to an edge segment by using an asymptotic approximation. Our method is illustrated by showing simulations and real-image examples.

The robust statistics has recently been adopted by the computer vision community. Various robust approaches in the computer vision research have been proposed in the last decade for analyzing the image motion from the image sequence. Because of the frequent violation of the Gaussian assumption of the noise and the motion discontinuities due to multiple motions, the motion estimates based on the straightforward approaches such as the least squares estimator and the regularization often produces unsatisfactory result. Robust estimation is a promising approach to deal with these problems because it recovers the intrinsic characteristics of the original data with the reduced sensitivity to the contamination. Several previous works exist and report some isolated results, but there has been no comprehensive analysis. In this paper robust approaches to the optical flow estimation based on the maximum likelihood estimators are proposed. To evaluate the performance of the M-estimators for estimating the optical flow, comparative studies are conducted for every possible combinations of the parameters of three types of M-estimators, two types of residuals, two methods of scale estimate, and two types of starting values. Comparative studies on synthetic data show the superiority of the M-estimator of redescending ψ-function using the starting value of least absolute residuals estimator using Huber scale iteration, in comparison with the other M-estimators and least squares estimator. Experimental results from the real image experiments also confirm that the proposed combinations of the M-estimators handle the contaminated data effectively and produce the better estimates than the least squares estimator or the least absolute residuals estimator.

An image obtained by ultrasonic medical equipment is poor in quality because of speckle noise, that is caused by the quality of ultrasonic beam and so on. Thus, it is very difficult to detect internal organs or the diseased tissues from a medical ultrasonic image by the processing, which is used only gray-scale of the image. To analyze the ultrasonic image, it is necessary to use not only gray-scale but also appropriate statistical character. In this paper, we suggest a new method to extract regions of internal organs from an ultrasonic image by the discrimination function. The discrimination function is based on gray-scale and statistical characters of the image. This function is determined by using parameters of the multi-dimensional autoregressive model.

We propose the Kernel MUSIC algorithm as an improvement over the conventional MUSIC algorithm. This algorithm is based on the orthogonality between the image and kernel space of an Hermitian mapping constructed from the received data. Spatial smoothing, needed to apply the MUSIC algorithm to coherent signals, is interpreted as constructing procedure of the Hermitian mapping into the subspace spanned by the constituent vectors of the received data. We also propose a new spatial smoothing technique which can remove the redundancy included in the image space of the mapping and discuss that the removal of redundancy is essential for improvement of resolution. By computer simulation, we show advantages of the Kernel MUSIC algorithm over the conventional one, that is, the reduction of processing time and improvement of resolution. Finally, we apply the Kernel MUSIC algorithm to the Laser Microvision, an optical misroscope we are developing, and verify that this algorithm has about two times higher resolution than that of the Fourier transform method.

2D (two-dimensional) convolution is a basic operation in image processing and requires intensive computation. Although the SIMD model is considered suitable for 2D convolution, previous 2D convolution algorithms on the SIMD model assume unbounded number of PEs (Processing Elements) available, which we call unbounded case. Unbounded case could not be satisfied on real computers. In this paper, time-optimal data-parallel 2D convolution is studied on mesh-connected SIMD computers with bounded number of PEs. Because the optimal computation complexity is not difficult to achieve, the main concern of this paper is how to achieve optimal communication complexity. Firstly the lower bound computation complexity is analyzed. Then the lower bound communication complexities are analyzed under two typical data-distribution strategies: block-mapping and cyclic-mapping. Based on the analysis result, an optimal algorithm is presented under the block-mapping. The algorithm achieves the lower bound complexity both in computation and in communication.

Intending to contribute to constructing better multimedia network systems, we propose a new concept of image database system of which form of storage is featuring exponential or graceful oblivion and abrupt recollection like the human memory property. By virtue of this property of database storages that is realized by employing hierarchical or pyramidal image coding, the database memory and transmission costs can be significantly reduced. In this paper we will describe the details of the concept, the results of theoretical analysis based on a simplified model which reveals the effectiveness of the proposed system, the structure of an experimental prototype system and the result of an experimental image retrieval service carried out by implementing it over ATM high speed channels.

The transform coding scheme is often used for data compression of images, but the blocking effects peculiar to the scheme appear more clearly in reproduced images as a coding rate (bits/pixel) decreases. These effects can sometimes be viewed as a periodical square-grid overlaying the images. In this paper,we propose a new method for selectively measuring the above blocking effects among several types of image degradation by means of the techniques of nonlinear signal processing for spectral infomation (cepstral techniques), in order to compare the amount of blocking effects for the different coding images. First a two-component model which consists of DC and AC images, is discussed from a viewpoint of subimage-by-subimage coding, and some basic properties of cepstral information for the model are investigated. Then we show a procedure to compute the cepstral information for two-dimensional image signals taking the horizontal and vertical directions ioto account, and introduce a cepstral mean square error (CMSE) as a new measure to estimate the amount of blocking effects. The computer simulation results for some test images using different coding schemes show that the amount of blocking effects in each image can be easily measured and estimated by this method even when the blocking effects appear slightly.

In retrieving information from databases widely distributed in a network, the first thing to do is to search and find the database where the required information is stored. We call this the information searches rather than the retrievals. In this paper, we present a search and retrieval method for multimedia information, especially images. First, we formalize the general elements of information search and introduce a new search concept based on entropy reduction. Next, we discuss recent new technologies for image retrieval and introduce a new image retrieval system called VideoReality. Third, we present several methods of searching in the network- for example, the Internet robot TITAN, and a new search method for images distributed in the network that is based on the hierarchical structure of image retrieval. Finally, we discuss the network control and design concepts appropriate for information search and retrival.

The medial axis transform (MAT) is an image representation scheme. For a binary image, the MAT is defined as a set of upright maximal squares which consist of pixels of value l entirely. The MAT plays an important role in image understanding. This paper presents a parallel algorithm for computing the MAT of an n n binary image. We show that the algorithm can be performed in O(log n) time using n2/log n processors on the EREW PRAM and in O(log log n) time using n2/log log n processors on the common CRCW PRAM. We also show that the algorithm can be performed in O(n2/p2 + n) time on a p p mesh and in O(n2/p2 + (n log p)/p) time on a p2 processor hypercube (for 1 p n). The algorithm is cost optimal on the PRAMs, on the mesh (for 1 p n) and on the hypercube (for 1 p n/log n).

In this work, a new structure of M-channel linear-phase paraunitary filter banks is proposed, where M is even. Our proposed structure can be regarded as a modification of the conventional generalized linear-phase lapped orthogonal transforms (GenLOT) based on the discrete cosine transform (DCT). The main purpose of this work is to overcome the limitation of the conventional DCT-based GenLOT, and improve the performance of the fast implementation. It is shown that our proposed fast GenLOT is superior to that of the conventional technique in terms of the coding gain. This work also provides a recursive initialization design procedure so as to avoid insignificant local-minimum solutions in the non-linear optimization processes. In order to verify the significance of our proposed method, several design examples are given. Furthermore, it is shown that the fast implementation can be used to construct M-band linear-phase orthonormal wavelets with regularity.

A hierarchical relaxation method is presented for detecting local features in moving images. The relaxation processes are performed on the temporal-spatial pyramid, which is a multi-resolution data structure for the moving images. The accurate and high speed edge detection can be obtained by using infomation in the neighboring frames as well as the processed results in the higher layers of the pyramid.