Appendixes I Spatial Filtration for Video Line Removal
Total Page:16
File Type:pdf, Size:1020Kb
Appendixes I Spatial Filtration for Video Line Removal GORDON W. ELLIS The growing popularity in microscopy of video recording and image-processing techniques presents users with a problem that is inherent in the method-horizontal scan lines. These lines on the monitor screen can be an obtrusive distraction in photographs of the final video image. Care and understanding in making the original photograph can minimize the contrast of these lines. Two simple, but essential, rules for photography of video images are: (l) Use exposures that are multiples of the video frame time (l/30 sec in the USA). An exposure time less than this value will not record a completely interlaced image.* (2) Adjust the v HOLD control on the monitor so that the two fields that make up the frame are evenly interlaced. Alternate scan lines should be centered with respect to their neighbors (a magnifier is helpful here). t Following these rules will often result in pictures in which the scan lines are acceptably unobtrusive without recourse to further processing. If the subject matter is such that the remaining line contrast is disturbing, Inoue (l981b) has described a simple technique that can often yield satisfactory results using a Ronchi grating. However, on occasion, when important image details are near the dimensions of the scan lines, the slight loss in vertical resolution resulting from this diffraction-smoothing method may make it worth the effort to remove the lines by spatial filtration. The technique of spatial filtration, pioneered by Marechal, is described in many current optics texts. A good practical discussion of these techniques is found in Shulman (1970). To produce the pictures for the present article, I used a somewhpt simpler apparatus than the optical correlator illustrated in Shulman (see also Walker, 1982). In the present instance, the apparatus consists of: a monochromatic light source; two surplus aerial camera lenses (Kodak Aero-Ektar 12-inch f/~.5); an enlarger's negative carrier; and a 35-mm single-lens reflex camera with an extension bellows and a long-focal-length lens (in this case an ancient 7-inch Kodak Rapid-Rectilinear objective borrowed from a Kodak Model 3A Folding Pocket Camera, ca. 1910). The camera must have a plain ground-glass focusing screen with a clear central disk bearing a cross for parallax checking. I use a Nikon F with a type "C" screen. The light source for the optical processor is a vibrating light-fiber conveying light from an argon ion laser (Ellis, 1979). The source does not need to be a laser, but the requirements are easily met by the laser source described and it was available. The requirements are: high intensity, monochromaticity, and small size (but not too small). A 100-watt HBO 100 superpressure mercury arc has about the right source size. If fitted with a heat filter and a 546-nm interference filter and *See also page 426 regarding minimum suggested exposure time. tSee footnote *on page 256. GORDON W. ELLIS • Department of Biology, University of Pennsylvania, Philadelphia, Pennsylvania 19104 463 464 APPENDIX I housed in a suitable box (for safety as well as stray light control) without a lens, it would provide a more economical source for a dedicated optical processor. The two Aero-Ektars are mounted on a common axis with the light source and the camera. The light source is at the entrance pupil (film plane) of the first Aero-Ektar, which is face to face with its mate and about 13 em away. The negative carrier is mounted between them about 10 em from the second lens. Ideally, these lenses would be mounted with their front focal planes coinciding with each other and the film carrier, but this is not a necessary condition for this purpose. The closer spacing was dictated by space limitations. The spatial filter is mounted on axis at the back focal plane of the second Aero-Ektar and is carried on the barrel of the 35-mm camera's lens. The camera is focused on the object transparency in the negative carrier. The action of the spatial filter is selectively to prevent the light diffracted by the scan lines in the transparency from reaching the camera film plane where they would produce line images by interference with the undiffracted zeroth order. The simplest spatial filter that would eliminate the lines (as in the classic Abbe-Porter experiments) is an iris diaphragm closed down to the point that the first orders of the line diffraction pattern are just blocked. At first glance, this might seem the optimum filter for this purpose; however, video camera modulation transfer functions are not necessarily the same for vertical and horizontal detail. In fact, today's instrumentation cameras typically offer higher resolution in the horizontal direction (parallel to the scan lines) than can be achieved vertically within the constraints of the standard (USA) 525-line format. Consequently, in front of the camera )ens , I use a glass plate bearing opaque spots about 50% larger than the light source image and spaced to occlude all the diffracted orders from the scan lines while leaving the zeroth order and most of the aperture unobstructed. Then, to reduce stray light, I close the lens iris down to the level of the second orders of the diffraction pattern. Two accessories simplify adjusting the processor. They are an auxiliary focusing magnifier (actually a telescope), used to aid parallax focusing, and an aperture-viewing magnifier. The latter is FIGURE 1-1 . Original. SPATIAL FILTRATION FOR VIDEO LINE REMOVAL 465 used to examine the exit pupil of the camera's viewfinder, which provides a clear and magnified image of the filter and the diffracted light that it must be positioned to block. Photographically, there are several options in going from the original lined photograph to the filtered negative for the final print. One could, in principle, treat the optical processor as a special giant enlarger and project unlined prints from lined negatives. Unfortunately, this would move the printing easel well into the next room, in addition to requiring impractically long exposures. Otherwise, the shortest path in steps, though not necessarily in time, is to photograph the spatially filtered image on S0-185 Rapid Processing Copy film (EKC). Because this a self-reversing film, the developed S0-185 image is a filtered negative that can be used in an ordinary enlarger. I have found that this works well using the 488-nm line of the argon ion laser, but is about 10,000 times slower using the 546-nm light from a 100-W mercury arc, and will not work at all with a helium neon laser. An equally direct alternative I have not yet tried is to use S0-185 to make the original photograph of the video monitor. The resulting positive transparency could then be filtered in the optical processor and photographed on Panatomic-X, Plus-X, or other panchromatic film to produce the filtered negative. Using a CRT with a highly actinic phosphor might make this a practical approach. Conventional monitors would probably require excessively long exposures. For the photographs shown, I have followed a more complex path, but one that allows use of panchromatic film and short exposures. The original negative was exposed on Pan-X at 1/30 sec. This negative was copied onto Kodak Technical pan and developed in I : 3 Microdoi-X to produce a long-scale positive transparency for use in the processor. The filtered negative was then made on Pan-X. Here the exposure was 1/ 125 sec using 0.5 W from the 514-nm argon laser. Figure I-1 is a print from the original negative. The photograph does not show optimal adjustment of the interlace, but instead represents a less favorable case to demonstrate the power of the filtering technique. The result, the spatially filtered print in Fig. I-2, is entirely free of the video scan lines. Note the subjective impression that the delined print looks sharper than the original. FIGURE 1-2. Spatially filtered reproduction. 466 APPENDIX I The subject of the video micrograph is a diatom frustule that is one of 50 on an arranged slide from Turtox (No. Bl.434). It is viewed in polarized light through a 43/0.65-NA Leitz stress-free achromat and displayed on a 9-inch Panasonic monochrome monitor by a Dage-MTI Model 65 camera equipped with a Newvicon camera tube. Magnification of the image on the monitor was 1980X and covers 93 IJ.m horizontally, edge to edge. II Modulation Transfer Function Analysis in Video Microscopy ERIC W. HANSEN 11.1. INTRODUCTION The modulation transfer function (MTF) is a powerful tool for the analysis of imaging systems. It describes in quantitative terms the relationship of object and image, and provides a convenient framework for analyzing the performance of complex imaging systems, which may combine optical and electronic components. This appendix first presents a brief tutorial introduction to the MTF, and then illustrates its use with an analysis of a typical video microscope system. 11.2. MODULATION TRANSFER To begin, let us consider the imaging system shown schematically in Fig. II-1. An object with a sinusoidal spatial variation of intensity, / 0 (x0 ) along the X0 axis, is imaged with unity magnifica tion, producinga sinusoidal image intensity 1/x). (The subscripts o and i denote object and image, respectively.) The peak-to-peak distance along the sinusoid* is its spatial period; the reciprocal of the period is the spatial frequency,Jx, expressed in units of cycles per unit length (e.g., cycles/mm). In general, we may observe three things about the image.