SCANNING Vol. 12,141-146 (1990) OFACMS, Inc. Received December 19, 1989 The Utility of an On-Line Digital Image Recording System for SEM E. OHO, K. KANAYA Department of Electrical Engineering, Kogakuin University, Tokyo, Japan Summary: An on-line digital image processing system provides several useful functions to SEM users. It is especially convenient for general SEM image storing and filing, preferable to conventional methods using video monitor, camera, and film. When equipped with state-of-the-art magneto-optical disk unit (external memory) and a video copy processor for printout of a digital SEM image, its usefulness is enhanced. In order to validate this useful function, we compare a modern on-line digital image recording system with a conventional one, from the viewpoint of the general SEM user. Introduction Most SEM users utilize a conventional recording system consisting of a video monitor, with a resolution of 1000-2000 lines, and a high-performance camera. It is generally believed that the system is satisfactory for the average SEM user. In fact, the performance of the conventional recording system sometimes fails to match the high quality (e.g., high signal to noise as well as the high resolution) SEM images. Oho et al. (1986) have already confirmed this situation by comparing it with an on-line digital recording system which is closer to the ideal for SEM images (SEM images are essentially the electric signal). Moreover, the intrinsic ability of commercial SEMs (e.g., electron gun and lenses) has been improving. Address for reprints: Eisaka Oho Department of Electrical Engineering Kogakuin University 1-24-2, Nishishinjuku Shinjuku-ku, Tokyo 163-91, Japan SEM users go to great lengths to develop and print films and file data by conventional means. A digital image processing system, on the other hand, is a very convenient system for SEM image recording and filing. The most original up-to-date techniques are not requisite to operate an on-line digital recording system. This kind of system is now well established by several SEM manufacturers, simple to operate, and user friendly. A low-cost system for digital image recording and processing has been presented by Desai and Reimer (1 985). However, on-line digital image recording is not yet standard in general SEM usage. This situation may be due to an undeserved-or an underevaluation-of the system. Commercial systems of this type have been introduced into general SEM usage, but have not found practical applications because the quality of the digitized image was inferior to that of conventionally recorded images. In addition, there was little to attract users regarding price, ease of use, processing speed, and so on. However, this situation recently changed with the appearance of a high-performance system equipped with useful accessories for SEM image recording, and the advantages of the on-line recording system can now be exploited. In the present report, we describe an online digital recording system from the standpoint of the general SEM user. An On-Line Digital Recording System for General SEM Users Figure 1 shows an on-line digital image recording and processing system (modified Nireco LUZEX 111) linked to a field emission SEM (Hitachi S-800). The SEM signal is digitized in an analog-to-digital converter (ADC) with maximum 2048 X 12 bit resolution. At the same time, data are stored in image memory (2048 x 2048 x 16 bit 2-frame equipment). However, since we often use two different recording modes with 1024 x 1024 X 8-bit (mentioned later), that memory is equal to 1024 X 1024 x 8-bit 16 frames. The content of the memory can be displayed on a video monitor (1024 x
142 Scanning Vol. 12, 3 (1990) 1024 pixels, 256 grey levels). In addition to the normal recording time (in our case 80 s), we can store images in various recording times (TV rate up to several hundred s/frame). This processing can be performed in real time. Rigorous considerations of the interfacing of a computer to a S(T)EM beyond the object of the present study have been discussed elsewhere (e.g., Statham 1982, Zubin and Wiggins 1980). External memories, a magneto-optical disk (512 MB), hard disk (40 MB), and floppy disk (1 MB x 2) are connected with the main frame. The magneto-optical disk (Sony NWP-539S, Fig. 1, item 1) which is necessary because the disk not only has a reading and writing speed very similar to a hard disk, but also a much larger memory capacity. In addition, the magneto-optical disk is easily interchangeable with other storage devices, such as the floppy disk. At present, each disk is comfortably operated in Microsoft disk operating system (MS-DOS), using friendly software for supporting disk operations. Another requisite is a method for printing the digitized SEM image. Our system utilizes a modern video copy processor using a thermal head (Mitsubishi SCT- P75, Fig. 1, item 2). Compared with a conventional FIG, 1 On-line digital image recording (2048 x 2048 pixels, 12- instant camera, this processor has a higher image qualibit precision) and processing system equipped with a magneto- ty1 lower running cost (20 mnts/image), higher prooptical disk unit and video copy processor. cessing speed (23 s/image), and is easy to use. As an FIG 2 A printout image (5 12 X 5 12 pixels, 64 grey levels) from the video copy processor.
E. Oho and K. Kanaya: Utility of on-line digital image recording 143 example of the printout image (512 x 512 pixels, 64 grey levels), an SEM image of a biological sample is shown in Figure 2. Figures 3-5 are photographs of the video monitor of the computer. Using the digitally stored image, processing and measurement can also be performed in high speed. For instance, two-dimensional fast Fourier transforms (FFT) with 32-bit calculation precision and various 3 X 3 space filters of 1024 x 1024 pixels can be performed in approximately 10 s and 1 s, respectively, owing to utilization of several high-performance digital signal processors (DSPs). These functions (use of the im- proved homomorphic filter using the FFT proposed by Oho et al. 1987) can be used by SEM users to record SEM images with very wide dynamic ranges, which the narrow range of fidelity of film does not allow. Users in some fields may be plagued with SEM images of this kind. Since a system like that which we consider here receives the image signal only from the SEM, we can easily link to existing SEMs without reconfiguring the system. The price of such a system with a satisfactory performance is about one quarter of that of an SEM. If the system function consists only of image recording and FIG. 3 Comparison of the difference in image quality between digitized SEM images with 2048 x 2048 and 51 2 x 512 pixels: (a) 2048 x 2048 pixels; (b) 512 x 512 pixels. These images are enlarged from a small area to show the difference clearly.
144 Scanning Vol. 12,3 ( 1990) FIG. 4 Recording mode using the average of two lines to reduce the number of scanning lines (from 2048 to 1024 lines): (a) digitized SEM image using the present mode; (b) conventional SEM image recordedon film. There is little visual difference between these images. storing, this fraction would be substantially reduced. It should be noted that, until a few years ago, the price of a digital image storage system with much lower performance was roughly equal to that of the SEM. Further price reductions will be expected in the near future. Comparison of the Difference in Image Quality Between Digitized SEM Images with 2048 x 2048 and 512 x 512 Pixels The SEM which we use produces the signal (information) by scanning with 2048 lines. Hence, the present on-line system can record all signals obtained from the SEM. However, most earlier commercial digital image storage systems (and especially the low-cost ones) are based on 512 X 512 pixel recording and processing. Hence, usually the SEM signal had to be transformed to a digitized SEM image with 5 12 x 5 12 pixels resolution by a one in four reduction. Most SEM users should find this image impractical in many instances, as is shown in the following. Figures 3a and b are digitized SEM images of an Au-coated LSI consisting of 2048 x 2048 and 5 12 x 5 12 (use of one in four lines) pixels, respectively. These photographically enlarged images show the difference clearly. The 512 X 5 12 image was recorded using the antialias filter (a lowpass filter) based on the well-known sampling theorem for reducing the artifacts in the sampled image. Not surprisingly, the image quality of Figure 3b (e.g., sharpness) is severely degraded, with a considerable loss of information. Most SEM users would not be pleased with this 512 X 512 image. They are accustomed to seeing conventional, high quality, SEM images with high resolution. Incidentally, if a certain blurred image (e.g., an image taken at a much higher magnification) is digitized, the 5 12 x 5 12 pixels would be adequate based on the sampling theorem, the details of which, as related to our field, have been described by Crewe (1 980) and Crewe and Ohtsuki ( 198 1). Simultaneous Dual Recording Modes In our system, display and processing are performed at 1024 X 1024 pixel resolution. The employment of a 2048 X 2048 display and processing system with a satisfactory performance is too costly for most SEM
E. Oho and K. Kanaya: Utility of on-line digital image recording 145 FIG. 5 High-resolution recording mode without thinning out scanning lines: (a) digitized SEM image; (b) conventional SEM image together with the extremely enlarged versions (a') and (b'). It should be noted that the image degradation caused by conventional recording system is very severe. users. However, we utilize two SEM imaging modes for recording with 1024 X 1024 pixels. One mode records without thinning out the scanning lines (we call it the high-resolution mode). This mode can memorize the information from roughly haif of the conventionally recorded area because the distortion at the edge of the SEM image field is not generally recorded on film. Of course, important data (e.g., SEM operating conditions, characteristics of a specimen, and so on) can be easily stored together with its image. We discuss the superiority of the quality of this image in the next section. When it is not very necessary to enlarge SEM images, another recording mode is effective. This mode uses the average of two lines calculated by a real-time arithmetic unit to avoid any loss of information due to thinning out lines and to increase the SNR of a SEM image. By employing this mode, we can obtain a digitized SEM image with satisfactory quality from the whole area. In order to confirm the ability of this mode,
146 Scanning Vol. 12,3 ( 1990) we recorded the SEM image of an LSI with fine details simultaneously in the image memory (Fig. 4a) and on film (Fig. 4b). There is little difference visually between Figure 4a and b. This mode may be effective for routine work. Confirmation of the Superiority in Image Quality of the On-Line Digital Recording System We compare the difference in quality of on-line digital versus conventional recording of the SEM image. The micrographs in Figures 5a and b are digitized (high-resolution mode) and conventional SEM images, respectively. The micrographs shown in Figures 5a' and b' are extremely enlarged images for Figures 5a and b obtained by a digital interpolation method using the FFT consisting of the hardware and a darkroom enlarger, respectively. Here, this interpolation method is equivalent to the cubic convolution method based on the sampling theorem (Bernstein 1976), which is one of the most powerful interpolation methods for digital images. We chose it for its ease of use and advantage of processing time in our system, compared with the cubic convolution method. Although surface structures are visible in Figure 5a, those in 5b are disturbed by the film-grain noise and blur, which may originate from nonideal point spread function in the conventional recording system. It should be noted that the image degradation caused by conventional systems is more severe than expected. The validity of structures in digitized images can be confirmed by observing an image (identical view of Fig. 5a) recorded at a much higher magnification than that of Figure 5a (Oh0 er al. 1986). Conclusions It has been demonstrated that an on-line digital recording system has advantages in image quality for recorded SEM images over the conventional recording system. In addition, the on-line system equipped with a magneto-optical disk controlled by friendly software for disk operation and a modern video copy processor is really practical for general SEM usage. Acknowledgment This work was supported by a Grant for Scientific Research from the Ministry of Education, Science and Culture of Japan. References Bernstein R: Digital image processing of earth observation sensor data. I5M J Res Dev 20,40-57 (1976) Crewe AV: Imaging in scanning microscopes. Ultramicroscopy 5, 131-138 (1980) Crewe AV, Ohtsuki M: Optimal scanning and image processing with the STEM. Ultramicroscopy 7,13-18 (1981) Desai V, Reimer L: Digital image recording and processing using an Apple I1 microcomputer. Scanning 7, 185-197 (1985) Oho E, Sasaki T, Kanaya K: A comparison of on-line digital recording with conventional photographic recording for scanning electron microscopy. J Etectron Microsc Tech 4, 157-162 (1986) Oho E, Sasaki T, Ogihara A, Kanaya K: An improvement in digital homomorphic filtering and its practical application to SEM images. Scanning 9, 173-176 (1987) Statham PJ: Interfacing of computers to a STEM. Ultramicros- COPY 8,309-320 (1982) Zubin JA, Wiggins JW: Image accumulation, storage, and display system for a scanning transmission electron microscope. Rev Sci Znstrum 5 1, 123-131 (1980)