PAPER A 4K 2 K-Pixel Color Image Pickup System

Transcription

1 IEICE TRANS. INF. & SYST., VOL.E82 D, NO.8 AUGUST PAPER A 4K 2 K-Pixel Color Image Pickup System Kohji MITANI, Hiroshi SHIMAMOTO, Nonmembers, and Yoshihiro FUJITA, Member SUMMARY We have developed an experimental 4 K 2K pixel progressive scan color camera system. This new camera system has a data rate of 297 MHzpixel/sec and 60 frame/sec and we are sure that horizontal and vertical limiting resolution of 1500 TVL (TV lines) can be achieved on a color monitor. Instead of the previous approach of improving resolution simply by increasing the pixel count in a imager, a novel four-sensor pickup method with 2/3 inch 2 million pixel CMD (Charge Modulation Device) imagers is used in this system. These sensors have 1920 (H) 1035 (V) pixels within a 16:9 wide aspect image area and are successfully driven at 148 M pixel/sec in the progressive scan mode. In the four-sensor pickup method, two sensors are used for green and the rest are for red and blue. A spatial offset imaging method in the diagonal direction was applied to the two green sensors to improve the horizontal and vertical resolution effectively. The horizontal and vertical resolution of the red and blue signals become half that of the green signal, because only one 2 M-pixel imager is used for each signal. The resolution of this system, however, is not degraded so much because the luminance signal is mainly composed of green signals. key words: very-high-resolution imaging, four-imager pickup method, CMD, HDTV 1. Introduction Paralleling advances in semiconductor processing technology, ever higher resolution solid-state imagers have been developed incorporating 8 million pixels and even up to 16 million pixels [1]. Yet operating at frame rates of only 2to 3 frames per second, these sensors have only been used in pickup systems for still images. Now HDTV (High Definition Television system) 2/3-inch 2- million pixel video imagers with 5 µm 5.2 µm pixels are being manufactured in volume and they can support frame rates of 30 frames per second [2], [3], but imagers offering even higher resolution have yet to be developed. Although images providing higher resolution than HDTV would apparently require smaller sized unit pixels than are currently available, reducing the size of pixels beyond a certain point invites general deterioration of sensitivity, dynamic range, signal-to-noise ratio (S/N), and other fundamental imaging attributes. Manuscript received May 13, Manuscript revised January 20, The authors are with three dimentional audio-visual system research division, Science and Technical Research Laboratories, NHK (Japan Broadcasting Corporation), Tokyo, Japan. The author is with Engineering Administration Department, NHK (Japan Broadcasting Corporation), Tokyo, Japan. What is more, such a very-high-resolution pickup system would also require a high-drive-frequency imager in order to sustain the real-time frame rates that are needed for video image acquisition. However, faster drive speed invites increased dark current due to the generation of heat in the imager, as well as increased noise from the surging high-speed drive pulses. This paper gives an overview of our recent work on a very-high-resolution imaging system that provides better resolution than HDTV cameras as a step toward a viable next-generation video pickup system. Although faster drive capability and increased pixel count will eventually be necessary to develop an image sensor for such a very-high-resolution pickup system, with today s technology it would be extremely difficult to implement a imaging system with a pixel count higher than the 2M-pixel imager. It was this dilemma that led us to propose an alternative approach of realizing enhanced resolution by increasing the number of imagers incorporated in the pickup system, thereby mitigating the burden on each constituent imager. This paper describes our implementation of an experimental 4 K 2K color video pickup system that accelerates the drive frequency of a newly-developed 2M-pixel Charge Modulation Device (CMD) and also achieves super-high resolution by adopting a novel fourimager pickup method. In Sect. 2of the paper we will give the specifications of the experimental system and provide an overview of the performance objective we are attempting to achieve with the imager. Section 3 will describe the characteristics of the CMD imager used in the experimental system, detailing the imaging performance of the imager when operated at high speed, and the level of noise associated with the broadband pre-amplifier. Section 4 will highlight the advantages of the multi-imager approach for application to high-resolution systems, and describe the configuration and performance of the four-imager scheme used in the present experimental pickup system. In Sect. 5 we will give our assessment of the system based on the output of imaging experiments conducted with the imager. We will conclude by summarizing the more significant findings reached through fabricating the system. 2. Experimental System Specifications Specifications of the experimental pickup system are

2 1220 IEICE TRANS. INF. & SYST., VOL.E82 D, NO.8 AUGUST 1999 summarized in Table 1. These specifications were based on the study of various technological issues increasing bandwidth to support enhanced resolution of larger screens, reduction in S/N, reduction of resolution in optical systems such as lenses and also with the idea that these specifications should provide a foundation for a viable next-generation imaging system that supports four times the picture data quantity of HDTV [4]. The basic system we propose has the same 16:9 aspect ratio as HDTV, but doubles the HDTV pixel count in the horizontal and vertical directions to 4 K 2K, or exactly, 3,840 pixels 2,070 lines. The system operates at a frame rate of 60 frames per second, and seeking to accommodate the image processing capability of computers, it operates in progressive scan mode. The basic frequency clock of the experimental system is set at 297 MHz (four times that of HDTV), and it also accommodates an HDTV image slicing capability to cut the images down from super high resolution to HDTV format. Currently no image sensor has been developed that satisfies these specifications. In this work in pursuit of higher resolution we have therefore integrated 2/3-inch 2M-pixel CMDs that were developed by Olympus Optical Corporation and a newly-conceived four-imager pickup scheme into an experimental video imaging system. Incorporating CMDs that can accommodate relatively higher speed operation, the experimental pickup system operates in progressive scan mode at data rates of 297 M pixel/sec. The four-imager method employs two imagers with their pixels offset diagonally for the green signal. Outputs from the two imagers are combined, and after some additional interpolation processing, the resolution of the green signal, which makes the greatest contribution to the overall luminance signal, increases. To display 2,000-line images we used the only 2,000-line color display monitor with a 1:1 aspect ratio that is commercially available at the present time. The imaging area of the pickup system has an aspect ratio of 16:9, so half the imaging area in the horizontal direction was masked off to display the images. The HDTV image slicing capability was used to display 1/4 of the imaging area on an HDTV monitor M-Pixel Progressive-Scan CMD Imager 3.1 Characteristics of CMDs [5] High resolution and fast drive capability were the two primary performance requirements we sought for the imager incorporated in the experimental pickup system. Image pickup tube and CCD (Charge Coupled Device) imagers are commonly used in image pickup systems, but they exhibit registration errors, diminished charge transfer efficiency when operated at high speed, generate too much heat, and have other deficiencies that make them a poor choice for our present purpose. CMD imagers, on the other hand, are a type of APS (Active Pixel Sensor) which means that they have one or more active transistors located within each pixel. CMDs have low power consumption, short reset time and other attributes making them well-adapted for high-speed drive operation. The primary concern with CMDs is that they exhibit fixed pattern noise (FPN) attributed to pixel-to-pixel variations, which necessitate FPN suppression circuitry implemented in the off-chip frame memory. However, this was not such a major obstacle in the present experimental system, because the FPN suppression circuitry was incorporated as part of a digital signal processor that had to be implemented anyway to integrate broadband signals. These were our primary reasons for selecting the CMD for the present system. Figure 1 shows a schematic diagram of the 2/3- inch 2M-pixel CMD used in this work. Each operation necessary for imaging accumulation, readout, reset, overflow, etc., is controlled by the voltage applied to the gate of each pixel. One will observe that vertical scanners are provided on both sides of the active image area, thus providing a structure that can be switched back and forth between interlaced and progressive scan modes. Interlaced scanning is achieved by making both Table 1 Specifications of the experimental pickup system. Fig. 1 Circuit configuration of 2/3 inch 2 M-pixel CMD.

3 MITANI et al: A 4 K 2 K-PIXEL COLOR IMAGE PICKUP SYSTEM 1221 vertical scanners active at the same time and continuously reading out the signals in two-line increments, while progressive scanning is made possible by only enabling one vertical scanner at a time. When applied to HDTV, the CMD operates in interlaced scan mode at a data rate of M pixel/sec, but in the present experimental system it operated in progressive scan mode at a data rate of M pixel/sec, double the HDTV data rate. We will now consider the imaging performance of the CMD when operated in progressive scan mode at high speed. 3.2Imaging Characteristics under High-Speed Drive Conditions [6] While operating the CMD at high speed, measurements were taken of the photo-conversion, increase in temperature, and the variation in FPN as a function of device temperature. The CMD used for the measurements was a 2/3-inch 2 M-pixel imager with two-line readout configuration. The horizontal scanner was driven at MHz when the CMD was operated in interlaced scan mode at HDTV rates. Figure 2shows the photo-conversion characteristics of the CMD when operated in interlaced scan mode at the HDTV rate and in progressive scan mode at two times the HDTV frequency. When operated in progressive scan mode, the signal output was half that when the CMD was operated in interlaced scan mode. This is because in progressive scan mode each individual pixel signal is outputted in contrast with the interlaced scan mode in which signals from two pixels are added and outputted at a time. Therefore even when driven at high speed, this did not present any major problems regarding the photo-conversion characteristics. The measured increase in CMD temperature and it s power consumption as a function of drive frequency, and the variation in FPN level as a function of CMD temperature is shown in Figs. 3 and 4, respectively. The CMD power consumption at MHz is 480 mw and this value is almost the same as that of 2M-Pixel IT- CCD [7] at MHz driving. This means that the power dissipation in CMD doesn t become a big problem at MHz and a simple solution can be used for reducing the sensor temperature. The CMD temperature was measured here by attaching a temperature sensor to the metal pad on the back side of the CMD. The CMD was driven at a frequency of MHz and its temperature rose about 25 C with respect to ambient temperature when no cooling was provided. When cooling with a Peltier device, however, the increase in CMD temperature lowered significantly, reaching only about 3 C. Figure 4 shows FPN peak-to-peak (p-p) values at a pre-amp output signal when placing the CMD in a constant temperature bath and driving it at MHz. At CMD temperatures below 60 C, FPN exhibits a mostly non-dependent relationship with temperature, but at temperatures above this value, it clearly becomes dependent. Such FPN is referred to as white spots and is thought to be caused by dark current in each pixel. From the above results, we can see that if the temperature within the camera head increases by 20 C above room temperature (20 C), the temperature of the non-cooled CMD will exceed 60 C Fig. 3 Increase of CMD temperature as a function of drive frequency. Fig. 2 Photo-conversion characteristics at interlaced and progressive scan modes. Fig. 4 FPN level as a function of CMD temperature.

4 1222 IEICE TRANS. INF. & SYST., VOL.E82 D, NO.8 AUGUST 1999 and the white spots will be conspicuous. While the white spots can generally be suppressed by the FPN suppression circuit since they are a form of FPN, they cannot be sufficiently suppressed in this way at temperatures above 60 C because the noise level dramatically changes with respect to CMD temperature. The experimental system is therefore equipped with a Peltier device to perform controlled cooling and suppression of white spots. 3.3 Pre-Amplifier Noise Just like a conventional pickup tube, the signal output from a CMD is output as current, and therefore has to converted to voltage using a transimpedance amplifier as a pre-amplifier. One advantage of the CMD is that, since the photodiode is reset through driftinduced charge sweep-out, very little random noise is generated in the device. In order to fully exploit this advantage, it is necessary to minimize the additional noise added by the pre-amplifier as much as possible. The equivalent input current noise density (J n )ofan ordinary transimpedance amplifier is approximated by J 2 n =( 4kT/R) 2 +(2πfCE n ) 2 (1) This is where R is the feedback resistance, 4kT/R is the thermal noise current density of R, f is the bandwidth, C is the sum of the output capacitance of CMD and the input capacitance of the transimpedance amplifier, and E n is the equivalent input voltage noise density of the transimpedance amplifier. It is apparent from (1) that random noise consists of two components: a flat component caused by feedback resistance thermal noise that is unrelated to the frequency, and a triangular noise component caused by amplifier thermal noise that increases in proportion to the frequency. Particularly in the case of current signal readout imagers such as CMDs, the triangular noise component is dominant, because these imagers have such enormous output capacity (on the order of 20 pf) compared to image pickup tubes. In the present experimental pickup system it is especially important to increase the bandwidth so that a flat characteristic can be obtained for the pre-amplifier of up to at least 37 MHz, because the CMD is driven at a high clock frequency of MHz. Reducing the triangular noise is also an important consideration for increasing the S/N. Figure 5 shows the measured input current noise density of a feedback amplifier and a commonbase amplifier employed as a transimpedance amplifier, assuming they are connected to the 22 pf output capacity of a CMD. With the feedback amplifier we were able to minimize the pre-amplifier equivalent noise resistance and thereby increase the S/N by deploying two J-FETs with high mutual conductance in parallel as initial stage FETs, and thereby reduce the current noise density far more than using the common-base amplifier. Fig. 5 Current noise density on transimpedance amplifiers with 22 pf as CMD output capacitor. However, a random noise figure in the signal output of a CMD has been reported of 4 na rms for a bandwidth of 3 MHz [8]. This is thought to be channel noise produced by the amplifier transistor in each pixel, and if one assumes that it exhibit flat characteristics, then the current noise density of the CMD output would become 2.3 pa/ Hz. It should be apparent based on the results up to now that the noise on the pre-amplifier side which is included in the video signal dominates, and this preamplifier noise must be reduced even further. Thus, a major issue for the future is the development of a transimpedance amplifier optimized for current output type imagers such as CMDs that not only maintains a high gain to prevent the feedback resistance thermal noise from becoming large, but which also avoids amplifier saturation as a result of the clock component included in the output signal. This amplifier should also realize a higher bandwidth. 4. Enhanced Resolution Using the Four-Imager Method In this section we describe the four-imager method that has been employed in the experimental pickup system, designed to provide higher resolution by increasing the number of imagers. We will then highlight the advantages of the four-imager approach for achieving highresolution video acquisition compared with the alternative method of using a single imager but increasing the pixel count. 4.1 Overview of the Four-Imager Pickup Method [9] We developed this method for the purpose of capturing high-resolution color video using an imager with relatively fewer pixels. Figure 6 shows a schematic diagram

5 MITANI et al: A 4 K 2 K-PIXEL COLOR IMAGE PICKUP SYSTEM 1223 Fig. 6 Schematic diagram of the new four-imager color-separation prism. of the new four-imager color-separation optics. Incident light is separated into four color components two greens, red, and blue (GGRB) and sent to their respective imagers. The separation into red, green, and blue has exactly the same characteristics as the threeimager pickup method used in commercial and broadcast video cameras, and the color reproduction characteristics are also identical to the conventional approach. The most significant departure from the three-imager pickup method is that a half mirrored beam splitter has been inserted in the optical path of the green light, thus separating it into two identical green-light components. Resolution enhancement is then achieved by spatially offsetting the pixels of the two green-light imagers. Since the pixels are offset using optical images of the same wavelength, this reduces the weakening of the pixel offset effect caused by the color aberration that occurs when just red, green, and blue pixels are offset in the three-imager pickup method (red and blue imagers are offset at a half-pixel pitch with respect to the green imager) and this makes it easier to obtain a high-resolution signal. This four-imager method effectively enhances resolution by enhancing the green signal, which makes the greatest overall contribution to the luminance signal. By adopting the four-imager approach in the present experimental pickup system, the equivalent of a4k 2K pixel signal can be derived from the green signal coming from the output of the two CMD imagers for green after being subjected to interpolation processing. Red and blue components are allocated to just a single 2M-pixel each, thus producing 2K 1 K signals. Figures 7 (a) and (b) show the imaging sampling patterns for each color channel, and the Nyquist domains for the color channels, respectively. By offsetting the pixels of the green signal imagers diagonally, the resolution is predominantly enhanced in the horizontal and vertical directions. This method of arranging spatial sampling points in a quincuncial pattern effectively extends horizontal and vertical resolution, which is well suited to the acute visual preferences of people. At the same time, however, the horizontal and vertical resolution of the red and blue signals become half that of the green signal. To suppress the resulting aliasing components in the red and blue signals, optical low-pass filters (LPF) are inserted immediately in front of the red and blue CMDs that reduce the response at the horizontal and vertical sampling frequencies (i.e., 2,160 horizontal TV lines and 2,070 vertical TV lines) to zero. The effect of performing spatial offsetting between two imagers are significantly influenced by the accuracy in mounting imagers to the prism. Figure 8 shows the relationship between the offset of the two imagers and the limiting resolution in the horizontal direction. The limiting resolution point is defined here as the spatial frequency when the power of the aliasing components and the power of the baseband signal coincide. The limiting resolution in the vertical direction behaves similarly, though this figure shows the variation of the only horizontal limiting resolution. As seen in the figure, a mounting accuracy within +/ 0.5 µm makes it possible to achieve a limiting resolution above 1800 TV lines. 4.2More Imagers or More Pixels? We observed earlier in Eq. (1) that if one attempts to fix the frame rate and increase the number of pixels in an X Y addressable imager like a CMD, this causes the output capacity of the readout line C and signal bandwidth f to increase. This pushes up the pre-amplifier noise precipitously, which causes the S/N of the signal output to deteriorate. In order to mitigate this amplifier-noise-induced S/N deterioration, it is necessary to divide up the output capacity by implementing multiple output signal lines, and to avoid increasing the bandwidth. In addition, reducing the size of a unit pixel to increase the pixel count causes the sensitivity, saturation signal level, and the S/N to deteriorate. Figure 9 shows the photo-carrier (hole) to signal current conversion ratio as a function of pixel size in the CMD [10] [12]. Although the pixels can be scaled down to about 5 µm 5.2 µm without much of a loss in detection sensitivity, the sensitivity can be expected to deteriorate sharply if the elements are shrunk even further. In addition, reducing the unit pixel size causes signal contamination from adjacent pixels and deterioration in the sensitivity to red light because the area separating pixels become narrower and the depth of the pixel becomes shallower. However, the four-imager method adopts a different approach to improved resolution. This approach employs multiple imagers with relatively fewer pixels, and derives a high-resolution video signal by offsetting the patterns of spatial sampling points of the optical image, then integrating the output signals. Although

6 1224 IEICE TRANS. INF. & SYST., VOL.E82 D, NO.8 AUGUST 1999 Fig. 7 Spatial offset imaging method and Nyquist domains for green, red, and blue channels. Fig. 8 Depedence of limiting resolution on the offset between the two green CMDs. Fig. 9 Photo-carrier (hole) to signal current conversion ratio as a function of pixel size in the CMD. the number of signal output lines must be increased to accommodate the multiple imagers, this is no different than having to increase the number of output lines in the increased-pixel-count approach to suppress the increased amplifier noise. Also, the incident light per unit pixel cell is reduced in the multi-imager approach because the optical image is split by a prism, but in the increased-pixel-count approach the light per pixel is reduced about the same because the size of the pixels is reduced. One major advantage of the multiimager method is that it uses imagers with relatively larger pixels, so it is able to realize far better imaging characteristics than the alternative approach of simply increasing the pixel count. Especially in the very-highresolution video domain, where the increased-pixelcount approach would require reducing the size of pixels beyond the critical 5 µm 5.2 µm, thereby triggering rapid deterioration of imaging characteristics, the proposed multi-imager method offers a clear-cut advantage. 5. Imaging Experiments 5.1 Experimental System Configuration Figure 10 shows a block diagram of the very-highresolution experimental image pickup system. Implementing the four-imager method, four HDTV 2/3-inch 2M-pixel CMDs with a 16:9 aspect ratio are mounted in the camera head. As discussed earlier, two of the CMDs are for green light (G1 and G2), while the other two CMDs are for red and blue light. Resolution is effectively enhanced in the horizontal and vertical directions by mounting a prism between the two green signal imagers that realizes a half-pixel pitch diagonal offset between the pixels of the two imagers. Then, in the signal processor, the output of the two green channel imagers is combined and missing pixels are interpolated from four adjacent pixels to produce a progressive scan video stream at 60 frames per second with a horizontal resolution of 3,840 pixels and a vertical resolu-

7 MITANI et al: A 4 K 2 K-PIXEL COLOR IMAGE PICKUP SYSTEM 1225 Fig. 10 Block diagram of the experimental image pickup system. tion of 2,070 pixels. After interpolation, the signal goes through further processing steps including gamma correction and image extraction. Finally, half the image data (i.e., 1,920 dots horizontal 2,070 lines vertical) are D/A converted and output to the monitor. The experimental system is also equipped with an HDTV image slicing capability that permits part of the acquired image to be displayed on an HDTV monitor. 5.2Imaging Experiments The relationship between the image acquisition area and the image display area is illustrated in Fig. 11. To display 2000-line images (actually 1,920 horizontal pixels 2,070 vertical lines) on a display monitor with a 1:1 aspect ratio, the vertical-to-horizontal symmetry of the images was 8-to-9, slightly longer in the horizontal direction. In the case of the HDTV sliced images, on the other hand, because a 16:9 aspect ratio part could be extracted for display on the HDTV monitor, they could be displayed with a symmetry of 1. The cutting position can be scrolled laterally or vertically one pixel at a time, so any portion of the image can be readily displayed. Figures 12(a) and (b) show reproduced pictures of HDTV retoma charts displayed on the 2,000-line monitor. From Fig. 12(a) it is apparent that only about half of the total lateral image data is displayed on the monitor. From the enlarged central portion of the im- Fig. 11 Relationship between image acquisition area and image display areas on 2 K-line monitor and HDTV monitor. age shown in Fig. 12(b), one can see that a limiting resolution of 1,500 TV lines has been achieved in both horizontal and vertical directions. This limiting resolution can be attributed to a number of factors: the number of pixels for the red and blue channels is fairly low, the high-frequency response is diminished by interpolation processing and because the apertures of pixels overlap in the pixel offset method. One should be able to address these problems fairly easily, say with an aperture enhancement circuit that electrically elevates

8 IEICE TRANS. INF. & SYST., VOL.E82 D, NO.8 AUGUST (a) (b) Fig. 12 Reproduced pictures of HDTV retoma chart displayed on the 2 K-line monitor. (a) 1,920 (H) pixels 2,070 (V) lines on a display monitor with a 1:1 aspect ratio. The vertical-to-horizontal symmetry of the images was 8-to-9, slightly longer in the horizontal direction. (b) Central portion of (a). Fig. 13 A typical image produced by the system. Given the very high resolution of the image, there is very little loss of resolution even when the image is enlarged. vices (CMDs). The experimental system not only fully exploits the inherent advantages of CMDs high resolution and high-speed drive capability but it also exhibits a viable method of enhancing resolution by implementing a spatial pixel offset method between two of the four CMD imagers for the green signal channels. A series of imaging experiments demonstrate that the system has a limiting resolution of 1,500 TV lines in both the horizontal and vertical directions. More fundamentally, this work shows that, in the super-high-resolution domain that goes beyond HDTV, the approach adopted here of increasing the number of imagers has inherent advantages, because it avoids the worsening of fundamental imaging attributes that tends to occur when the size of pixels is reduced beyond a certain point. Acknowledgments the high-frequency amplitude characteristics. A typical image produced by the system is shown in Fig. 13. Given the very high resolution of the image, there is very little loss of resolution even when the image is enlarged. Since there is currently no system available on which a 2,000-line image can be directly evaluated, we plan to follow up this work by trimming the images to HDTV format and subjecting them to quantitative assessment for the HDTV domain for which measurement equipment is available to improve the picture quality of these high-resolution images even more. 6. Conclusions This paper describes our implementation of a 4 K 2 K pixel experimental pickup system that operates in progressive scan mode at a rate of 60 frames per second. The system adopts a novel four-imager approach that employs four 2/3-inch 2 M-pixel charge modulation de- The authors would like to thank Mr. Tsutomu Nakamura, Mr. Kazuya Matsumoto, Mr. Tetsuo Nomoto and Mr. Hideharu Miyahara of Olympus Optical Co., Ltd., Integrated Precision Technology Department, for their kind assistance during the course of this work, especially in providing CMDs, advice on drive methods and useful discussion. References [1] J.R. Janesick, et al., New advancements in charge-coupled device technology: Subelectron noise and pixel CCDs, SPIE Proceedings, vol.1242, pp , Feb [2] K. Harada, M. Negishi, T. Ohgishi, S. Kubota, T. Oda, and M. Yamagishi, A 2/3 inch 2 million pixel FIT-CCD image sensor for HDTV camera system, ISSCC Dig. Tech. Papers, pp , Feb [3] Y. Toyoda, K. Itakura, T. Nobusada, Y. Saitoh, N. Kokusenya, R. Nagayoshi, H. Tanaka, M. Ozaki, M.

9 MITANI et al: A 4 K 2 K-PIXEL COLOR IMAGE PICKUP SYSTEM 1227 Sugawara, K. Mitani, and Y. Fujita, A 2/3-inch 2.0 M-pixel M-FIT CCD with a single channel HCCD for HDTV camera, ISSCC Dig. Tech. Papers, pp , Feb [4] SMPTE Standard, SMPTE 240M-1988, television signal parameters 1125/60 high definition production system, SMPTE J., vol.69, pp , Nov [5] T. Nomoto, R. Hyuga, S. Nakajima, I. Takayanagi, T. Isokawa, K. Matsumoto, and T. Nakamura, A 2/3-inch 2 M-pixel CMD image sensor with multi-scanning function, ISSCC Dig. Tech. Papers, pp , Feb [6] K. Mitani, H. Shimamoto, and Y. Fujita, A 2 K 2 K image acquisition system using four CMD imagers, SPIE Proceedings, vol.2663, pp.29 35, Jan [7] M. Morimoto, K. Orihara, N. Mutoh, K. Minami, K. Hatano, M. Furumiya, K. Arai, T. Nakano, Y. Kawakami, S. Kawai, I. Murakami, S. Suwazono, A. Tanabe, T. Tanaka, S. Katoh, Y. Urayama, A. Kohno, E. Takeuchi, N. Teranishi, and Y. Hokari, A 2/3-inch 2 M-pixel IT-CCD image sensor with individual p-wells for separate V-CCD and H-CCD formation, ISSCC Dig. Tech. Papers, pp , Feb [8] T. Nakamura, K. Matsumoto, and T. Nomoto, A CMD image sensor An approach to high-resolution imaging, ITE Tech. Rep., vol.19, no.25, pp.13 18, May [9] M. Sugawara, K. Mitani, T. Saitoh, T. Fujita, and K. Suetsugi, Four-Chip CCD camera for HDTV, SPIE Proceedings, vol.2173, pp , Feb [10] K. Matsumoto, I. Takayanagi, T. Nakamura, and R. Ohta, The operation mechanism of a charge modulation device (CMD) image sensor, IEEE Trans. Electron Devices, vol.38, no.5, pp , May [11] K. Matsumoto, I. Takayanagi, T. Nakamura, and R. Ohta, Analysis of operational speeds and scaling down the pixel size of a charge modulation device (CMD) image sensor, IEEE Trans. Electron Devices, vol.38, no.5, pp , May [12] M. Ogata, T. Nakamura, K. Matsumoto, R. Ohta, and R. Hyuga, A small pixel CMD image sensor, IEEE Trans. Electron Device, vol.38, no.5, pp , May Engineers. center of NHK. Hiroshi Shimamoto received the M.S. degree in electrical engineering from the Tokyo Institute of Technology in He joined NHK (Japan Broadcasting Corporation) in Since 1993, he has been with the NHK Science and Technical Research Laboratories, engaged in the research and development of high resolution and inteligent television camera systems. Mr. Shimamoto is a member of the Institute of Image Information and Television Yoshihiro Fujita received B.S., and Ph.D. degrees in electrical engineering from University of Tokyo in 1976 and 1998 respectively. Since 1976, he has been with NHK (Japan Broadcasting Corporation) Tokyo, Japan. He has worked on the development of various broadcast cameras including HDTV CCD cameras, super high-sensitivity cameras and highspeed cameras. Currently, he is a Senior Engineer at the engineering development Kohji Mitani received B.S. and M.S. degrees in electronic engineering in 1985 and 1987 respectively, and the Ph.D. degree in 1999, from Kyoto University. In 1987, he joined the Science and Technical Research Laboratories, NHK. Since then he has been working on solid-state image sensor camera. His research interests include very high data rate image pickup method, particulary a very high resolution and very high frame rate image aquisition system. He is a member of the Institute of Image Information and Television Engineers.