A hybrid scheme of subsampled DPCM and interpolative DPCM is proposed for the HDTV coding. A large bit rate reduction is realized by the two effects of pixel reduction and bit reduction. SN ratio is greatly improved compared with that of the coventional DPCM.

This paper presents a new scheme to estimate depth in a natural three-dimensional scene using a multi-viewpoint image set. In the conventional Multiple-Baseline Stereo (MBS) scheme for the image set, although errors of stereo matching are somewhat reduced by using multiple stereo pairs, the use of square blocks of fixed size sometimes causes false matching, especially, in that image area where occlusion occurs and that image area of small variance of brightness levels. In the proposed scheme, the reference image is segmented into regions which are capable of being arbitrarily shaped, and a depth value is estimated for each region. Also, by comparing the image generated by projection with the original image, depth values are newly estimated in a top-down manner. Then, the error of the previous depth value is detected, and it is corrected. The results of experiments show advantages of the proposed scheme over the MBS scheme.

The transient characteristics of the delay time of a signal are analyzed in detail in the time-varying cases of the continuously variable delay devices, whose delay time can be varied by a control signal. A supposed delay time which corresponds directly to the instantaneous control signal is defined as an instantaneous delay time (t) to make clear the essential function of the variable delay devices. In such devices, first (t) is converted into the actual delay time τ(t) and then the amplitude and frequency of the input signal (only the frequency in case of CCDs) are modulated by a modulating signal, -. Although the /τ conversion is essentially nonlinear, τ(t) is basically a lowpass-filtered version of (t) and hence τ(t) is nearly equal to (t) only if (t) is a slowly varying function of time. 1 generally holds and any τ(t) satisfying this constraint can be realized in the variable delay devices. Such time-variation of the delay time causes the time-axis transformation (time-axis compression and expansion), which is one of the very useful applications in the signal processing. It is shown that any time-axis transformation and its inverse transformation can be performed in principle by the variable delay devices. Finally, fundamental experiments on the time-axis transformation and its inverse transformation are carried out by using CCDs as the variable delay devices and the major theoretical results are varified experimentally.

A fractal image coding scheme using classified range regions is proposed. Two classes of range regions, shade and nonshade, are defined here, A shade range region is encoded by the average gray level, while a nonshade range region is encoded by IFS parameters. To obtain classified range regions, the two-stage block merging scheme is proposed. Each range region is produced by merging primitive square blocks. Shade range regions are obtained at the first stage, and from the rest of primitive blocks nonshade range regions are obtained at the second stage. Furthermore, for increasing the variety of region shape, the 8-directional block merging scheme is defined by extension of the 4-directional scheme. Also, two similar schemes for encoding region shapes, each corresponding to the 4-directional block merging scheme and the 8-directional block merging scheme, are proposed. From the results of simulation by using a test image, it was demonstrated that the variety of region shape allows large shade range regions to be extracted efficiently, and these large shade range regions are more effective in reduction of total amount of codebits with less increase of degradation of reconstructed image quality than large nonshade range regions. The 8-directional merging and coding scheme and the 4-directional scheme reveal almost the same coding performance, which is improved than that of the quad-tree partitioning scheme. Also, these two schemes achieve almost the same reconstructed image quality.

The Camera sensor network is a new advent of technology in which each sensor node can capture video signal, process and communicate with other nodes. We have investigated a dense node configuration. The requested processing task in this network is arbitrary view generation among nodes view. To avoid unnecessary communication between nodes in this network and to speed up the processing time, we propose a distributed processing architecture where the number of nodes sharing image data are optimized. Therefore, each sensor node processes part of the interpolation algorithm with local communication between sensor nodes. Two processing methods are used based on the image size shared. These two methods are F-DP (Fully image shared Distributed Processing) and P-DP (Partially image shared Distributed Processing). In this research, the network processing time has been theoretically analyzed for one user. The theoretical results are compatible with the experimental results. In addition, the performance of proposed DP methods were compared with Centralized Processing (CP). As a result, the best processing method for optimum number of nodes can be chosen based on (i) communication delay of the network, (ii) whether the network has one or more channels for communication among nodes and (iii) the processing ability of nodes.

The TAT system is an efficient analog bandwidth compression scheme for the HDTV. It compresses the bandwidth by reducing the pixels selectively from the non-detailed portions of the picture. The bandwidth can be reduced to half or less, while keeping high resolution in the detailed portions. In this paper, the basic concept, key techniques and some recent results of the TAT system are presented. First, it must be emphasized that the TAT system is a hybrid system of fixed subsampling and variable subsampling. The basic pixels obtained by the fixed subsampling transmit the basic structure of the picture. The additional pixels obtained by the variable subsampling transmit the details of the picture necessary for the HDTV. The concept of complementary subsampling is proposed and a restoration scheme of the removed high frequency component of the subsampled pixels is derived. Applying the concept to the TAT system, simultaneous transmission of an HDTV picture and an aliasing-free subsampled picture compatible with the current TV becomes possible. An efficient mode decision algorithm which determines the location of the additional pixels is developed. Time-varying fluctuation and partial degradation of resolution were observed in pictures having rich details. They are suppressed by introducing the control of mode transition and the interframe interpolation mode into the mode decision.

In this paper, we present an automatic vision-based traffic sign recognition system, which can detect and classify traffic signs at long distance under different lighting conditions. To realize this purpose, the traffic sign recognition is developed in an originally proposed dual-focal active camera system. In this system, a telephoto camera is equipped as an assistant of a wide angle camera. The telephoto camera can capture a high accuracy image for an object of interest in the view field of the wide angle camera. The image from the telephoto camera provides enough information for recognition when the accuracy of traffic sign is low from the wide angle camera. In the proposed system, the traffic sign detection and classification are processed separately for different images from the wide angle camera and telephoto camera. Besides, in order to detect traffic sign from complex background in different lighting conditions, we propose a type of color transformation which is invariant to light changing. This color transformation is conducted to highlight the pattern of traffic signs by reducing the complexity of background. Based on the color transformation, a multi-resolution detector with cascade mode is trained and used to locate traffic signs at low resolution in the image from the wide angle camera. After detection, the system actively captures a high accuracy image of each detected traffic sign by controlling the direction and exposure time of the telephoto camera based on the information from the wide angle camera. Moreover, in classification, a hierarchical classifier is constructed and used to recognize the detected traffic signs in the high accuracy image from the telephoto camera. Finally, based on the proposed system, a set of experiments in the domain of traffic sign recognition is presented. The experimental results demonstrate that the proposed system can effectively recognize traffic signs at low resolution in different lighting conditions.

The TAT system compresses the bandwidth of picture signals into about 1/2 by using variable subsampling with 2 or 3 subsampling modes. In this paper, a new variable subsampling scheme with many modes is proposed for higher compression. We call it the pyramidal TAT. In the proposed scheme, the basic 2-modes TAT is applied to the pyramidal representation of the picture. The pyramid used here is the complementary pyramid we proposed before. Variable subsampling with very many modes is easily done and a wide range of compression ratios are achieved with the pyramidal TAT. It is effective at compression ratios higher than 2 : 1 compared to the conventional TAT system.

We show that distributed source coding of multi-view images in camera sensor networks (CSNs) using adaptive modules can come close to the Slepian-Wolf bound. In a systematic scenario with limited node abilities, work by Slepian and Wolf suggest that it is possible to encode statistically dependent signals in a distributed manner to the same rate as with a system where the signals are jointly encoded. We considered three nodes (PN, CN and CNs), which are statistically depended. Different distributed architecture solutions are proposed based on a parent node and child node framework. A PN sends the whole image whereas a CNs/CN only partially, using an adaptive coding based on adaptive module-operation at a rate close to theoretical bound - H(CNs|PN)/H(CN|PN,CNs). CNs sends sub-sampled image and encodes the rest of image, however CN encodes all image. In other words, the proposed scheme allows independent encoding and jointly decoding of views. Experimental results show performance close to the information-theoretic limit. Furthermore, good performance of the proposed architecture with adaptive scheme shows significant improvement over previous work.