Aaron Williams. New Technologies Graphic Communication 302 Winter Quarter, 2003 Page 49

Transcription

1 Digital: the demise of film? Digital cameras are taking a bigger and bigger chunk of the consumer camera market every year. Camera companies are reaching mainly for this consumer-level crowd, but they are turning their focus more toward the professional user. This increase in digital camera use brings up some interesting questions. The important question is whether digital cameras can create the same image quality as film cameras. When it comes to imaging in the professional world, quality is a huge factor. Quality is most needed in the graphic arts and these are the types of jobs that have been paying the most attention to this long lasting debate, which may be nearing it s end. New technologies in digital cameras have greatly increased resolution over the last couple of years and are redefining some of the terminology that we are now accustomed to. As this evolution takes place, the important factors for print-quality images come under the microscope to determine whether film cameras still have an edge over digital cameras. Overall image quality in digital cameras is dependent on four things: dynamic range, sensitivity, signal-to-noise ratio, and resolution. Sensitivity is the amount of light that is collected by the photodiode. The more light that can be collected, the stronger the signal, and the better the picture quality. The signal-to-noise ratio is the amount of noise or distortion for a given signal. The higher the ratio, the lower the interference, which means better picture quality. Dynamic range is the range of tones, colors, and brightness that give a picture accuracy and depth. Image sensor dynamic range is a measurement of the sensor s ability to capture image detail across a range of dark to light areas in the image. The darker the darks and brighter the light areas that can be captured, the higher or better the dynamic range. While it s possible to directly compare a film s grain count with a pixel count, the number would be irrelevant because of the nature of the granules. Pixels have associated bits, called bit depth, that define color and grayscale. In a professional digital camera, each pixel can have anywhere from 36 to 48 bits of data to describe the state of a single pixel. A single frame of fine grain 1x1.5 35mm color film has an estimated 13 to 15 million individual silver halide granules. However, film is analog, not digital, so the state of each granule is, theoretically, infinitely variable. The combination of more tightly packed grains and infinite variability in each grain allows film to capture true continu- Aaron Williams New Technologies Graphic Communication 302 Winter Quarter, 2003 Page 49

2 ous-tone images, while digital always has a fixed number of steps or grayscale levels limited by the bit-depth of the pixel. With film cameras, tone rendition is determined primarily by choice of film and modified by filters, with further modification possible in the negative and printing. With digital, this is determined by the camera s response as modified by filters, with further modification in the camera and after download digital processing and printing. The sensor defines limiting resolution in digital cameras, which is almost always capable of less resolution than film cameras of similar cost and style. Analog film can, by its very nature, produce greater tonality, while some digital technology, by design and engineering, can capture an appreciably higher dynamic range. Because the photosensitivity of each granule is set according to its chemical composition and size, it has limits on how much light it can register and absorb. Too little light, and no photochemical reaction takes place. Too much light, and the photons overwhelm the granule and blow it out to a uniform monochromatic black. The dynamic range is expressed, not in decibels like an image sensor, but as a logarithm of those decibels, on a scale of 0-5. A medium grain film averages a dynamic range of about 4.0, depending upon the brand and emulsion. That translates into about 6-7 f-stops on a camera lens. Anything above or below is expressed as monochromatic black or white, with no detail whatever. Some digital camera image sensors are capable of capturing significantly more detail than film in the highlights and shadows. A typical high-end image sensor, such as the Philips 2x2K CCD found in many professional camera backs, can capture 11 or more stops of data; roughly double that of a comparable film stock. But, the extra dynamic range is useful only if it could be accurately reproduced on an output device. The gamut on film is much higher than digital, but they both come up against a seemingly inflexible bottleneck: the output device s gamut. For film, that s how many lines photographic paper is capable of resolving, and for pixels, it s the number of lines that can be reproduced by an ink jet printer or computer monitor. Both paper technologies are similar in terms of clay coating, brightness, opacity and gamut. The only way you can make use of film s much greater resolution is either by enlargement, a 35mm frame can be blown up to 14x17 or even 16x20 without an apparent loss of quality, or viewing it via transmitted rather than reflected light, as with a slide projector. The signal-to-noise ratio is also an important specification in image sensor capture quality. It is often represented in the description of a sensor s dynamic range. In other words, you cannot state Page 50 New Technologies Graphic Communication 302 Winter Quarter, 2003

3 the dynamic range of an image sensor without first knowing the noise it generates. Interestingly, image sensors need a larger dynamic range than film to emulate the high degree of detail that film can capture in both shadows and highlights. Image sensors are electronic devices with inherent uncertainties, inefficiencies and inaccuracies; all of which can result in unwanted artifacts or noise. Sensor size relative to the number of pixels can also affect noise. If you pack more pixels into the same size sensor the pixels get smaller and there is a greater probability of electronic interference, which creates noise, thereby reducing dynamic range. This is one reason that larger sensors are often preferred for higher quality digital cameras. Noise is a phenomenon encountered by all digital camera users at one time or another. The term noise describes a type of interference to which a digital image is subject. At its simplest level, a sensor counts photons striking its pixels as they reflect off a subject. This is the signal. However, one of the properties of a sensor is that no two pixels are likely to count the same number of photons, even when exposed to the same light source. That unpredictable variation in their counts is referred to as the background noise. Aside from this background noise, there are additional sources of noise that make their way into a digital image, such as readout noise and processing noise. These occur at each step of the process by which the signal created by the light striking a sensor is amplified and converted to a digital value. The noise is a combination of the random element that is part of the measurements with the addition of the interference caused by the camera s electronics. Even when there is no light striking the sensor, electrons accumulate gradually in its pixels, called dark noise, the signal it creates is indistinguishable from one produced by light. With digital cameras, noise tends to increase with two factors: sensor sensitivity and the length of the exposure. The noise content of a digital image increases along with the sensitivity gain of the sensor. As the sensitivity of the sensor is increased to light, so is the sensor s sensitivity increased to noise. For example, at 100 ISO, the noise content might be minimal, and only noticeable in areas where the light drops off in the image. The advantage of increasing the sensitivity of the ISO to obtain faster shutter speeds must be weighed against the increase in the noise of the image. The quantity of noise recorded in the image will vary greatly between different cameras. Taking photos at night can be a lot of fun. However, the low shutter speeds required also increase the likelihood of noise appearing in the images. The noise content can be controlled to some extent by the length of the exposure, and by the sensitivity of the sensor: a lower ISO setting will means a slower response from the sensor and in turn A CCD (Charge-Coupled Device) chip mounted on a camera circuit board. New Technologies Graphic Communication 302 Winter Quarter, 2003 Page 51

4 The CCD chip embedded in the camera body of a scientific digital camera body. a longer exposure; while a higher ISO setting will shorten exposure time. Deciding the best combination of these depends greatly on the specific camera used, as no two sensors are totally alike. Therefore, the best way to ascertain the settings that are the most noise prone, and which should be avoided, must be done by experimentation. The bigger the photosensitive area of a pixel, the more light it can collect and convert to electrons. But bigger pixels take up more space on a sensor, reducing the number of pixels that a given size sensor can hold. On the other hand, small pixels may allow for higher resolution on the same size sensor but their charge capacity is less, which means they are less photosensitive. Packing more pixels into the same size sensor also increases the potential for noise. Currently, the smallest pixel size for digital cameras is about 3.3 microns in a CCD and about 4 microns for CMOS. The popular Sony 3.3 MP sensor has a 3.45-micron pixel. Many other consumer digital cameras have pixels averaging around 5 microns. The new Kodak 14 MP CMOS has a 9- micron pixel, which delivers higher dynamic range, but the CCD itself is very large. Some other professional sensors have 12 micron pixels or even larger. Ever since digital SLRs came on the scene almost a decade ago manufacturers and photographers have shared two goals more pixels, and full frame coverage. In the late spring of 2002 the Contax N Digital was the first camera to give us full-frame coverage, but its digital capabilities were lacking in a number of areas. The Canon 1Ds is therefore really the first camera shipping that addresses professional photographer s need for a full-frame sensor. But now the question of pixel count needs to be addressed. The 1Ds has an 11 MP sensor. The forthcoming Kodak DCS 14n will have a 14 MP sensor. Are these enough? It looks like it, and here s why. Pixel count and sensor size are not interlinked. Consider consumer digicams. These cameras now have 5-megapixel sensors, not far short of the 6-megapixel sensor counts that are in DSLRs (Digital Single Lens Reflex cameras) like the Canon D60, Nikon D100 and Fuji S2. But the digicams put this large number of sensors into tiny chips, smaller than your smallest fingernail. Current DSLRs have sensors that are much larger, about the size of a 35mm film frame. The way that digicams accomplish this is by making the individual pixels smaller, about 3-4 microns in size versus the 6-10 micron size of the individual pixels in a DSLR. It could be assumed that manufacturers would simply be able to put more pixels into a DSLR imaging chip and we d end up with 20 or 30 MP fullframe cameras. But is it really necessary when you consider the fact Page 52 New Technologies Graphic Communication 302 Winter Quarter, 2003

5 that the Canon 1Ds, for example, with its 8.8-micron pixels and 11.1 MP resolution, is already capable of greater resolution than almost any 35mm format lens? This came to light when the first test reports started to appear, and every technical reviewer since has pointed out that the camera is able to record fine detail greater than standard resolution charts are able to display. Tests have shown that to really put this camera to the test, the best lenses at their optimum apertures are needed, otherwise the lenses let down the imaging chip. Small pixels have excellent resolution but suffer from increased noise, reduced exposure range, and reduced sensitivity. These effects are most noticeable on pixels smaller than 4 microns, which are used in most compact digital cameras. Large pixels have good SNR (signal-to-noise ratio), ISO speed and exposure range, but suffer from aliasing, low spatial frequency artifacts that appear when the lens has significant resolution. Aliasing usually shows itself as Moiré patterns on images with high frequency repetitive patterns, such as window screens and fabrics. It can be reduced by anti-aliasing (low pass) filters, which are expensive and unavoidably reduce resolution. Small sensors run into problems with lens diffraction, which limits image resolution at small apertures starting around f/16 for the 35mm format (43.3 mm diagonal). At large apertures, f/4 and above, resolution is limited by aberrations. There is a resolution sweet spot between the two limits, typically between f/5.6 and f/11 for good 35mm lenses. The optimum pixel size for high quality imaging seems to be in the 5-9 micron range. Larger pixels have problems with aliasing and can t take advantage of high quality lenses. Smaller pixels have more noise and less sensitivity, though they can still produce decent images. High-end cameras, like the Canon EOS 1Ds and Kodak DCS 14n, will stick with 5-9 micron pixels and evolve towards larger sensors with more pixels. A 24x36 mm sensor with 16+ megapixels (7.4 microns or less pixel spacing) is the holy grail of digital cameras. The performance of a camera like that will approach medium format, but it won t come cheap. Resolution refers to the number of pixels, both horizontally and vertically, used to either capture an image or display it. The higher the resolution the finer the image detail that can be seen. Resolution, as the word is used in digital imaging, is something of a misnomer. For purists who know photography and optics, resolution is the measure of the ability of a device to recognize individual converging lines, such as those on a resolution chart. But in the digital world another use of the word has prevailed ever since early monitors resolutions were described as the measure of the number of pixels that could be displayed. Resolution is the most talked about digital camera characteris- A CCD circuit is a massive array of microscopic phototransistors. New Technologies Graphic Communication 302 Winter Quarter, 2003 Page 53

6 tic and is often used to describe image quality. Yet, the lack of accurate information surrounding resolution makes it one of the most misunderstood discussion topics. Do all 3 megapixel cameras have the same pixel count? No, the large variety of terminology used is to blame for much of the confusion here. When a particular digital camera is referred to as having 3 megapixels, it may use a lot less than 3 megapixels to take the picture. There are many definitions of pixel count. The total pixels number refers to the overall number of pixels on the sensor, but this number is impertinent and misleading due to the fact that the image circle does not strike all pixels produced by the lens, nor are they used to capture the picture. Some of the pixels on the sensor are simply not used at all while others are used for picture processing, rather than picture capture. The pixel count in this picture area is called the picture pixels and is sometimes also referred to as image pixels. Since picture pixels are the only pixels used to capture the picture, the picture pixels number is a more accurate indication of a digital camera s intrinsic resolution than is the number of total pixels. When the image circle is smaller than that of the picture area of the sensor, the number of picture pixels is smaller than the total pixels number due to the un-maximized use of the picture area. This is one of the reasons why total pixels number can be misleading. The reverse, where the image circle is much larger than the picture area of the sensor, can also be true. Common examples of this include professional digital cameras such as Canon s D30 and EOS-1D as well as Nikon s D1 series cameras that use lenses designed for 35mm film photography. In these cameras, the picture area does not cover the full 35mm frame of 24 by 36 millimeters. The ratio of picture pixel area to the diagonal of 35mm film gives rise to the so called focal length multiplier. The number of effective pixels differs from picture pixels in that it also includes the pixels used to calibrate the black. This is accomplished by getting a ground or zero reading from masked pixels not exposed to light. The picture pixel information must be interpolated to create a viewable image. The number of pixels in this stored viewable image is the recorded pixel number. Most cameras have settings allowing users to store images of different sizes, essentially changing the number of recorded pixels. In certain instances, the number of pixels in the output image differs from that of the recorded pixels. Typically, this is due to cropping of the recorded image to fit standard dimensions such as VGA, SVGA, XGA, etc. or to fit aspect ratios such as 3:2, 4:3, and 16:9. Don t be fooled by the total pixel number. Rather, look at picture pixels or effective pixels, and output pixels to get a general idea of capabilities and limitations. However, take these numbers with a grain of salt. Page 54 New Technologies Graphic Communication 302 Winter Quarter, 2003

7 The pixel count, or resolution, of an image sensor has a direct relationship to the size image file that will be created. The more pixels, the larger the file is going to be. For instance, a VGA image sensor (with 640x480 active pixels) will produce approximately a 900K uncompressed file ( pixels times 3 bytes (R-G-B) per pixel = bytes). A 16 MP image sensor creates about a 48MB uncompressed file. It should be a very simple matter to count the number of pixels on an image sensor to determine the size picture it will capture. However, because of the many different numbers used by manufacturers to explain the resolution of their cameras, this can become difficult. The picture pixels are those that are used for the image capture. The picture pixels are the ones that are generally used when calculating the final uncompressed image size. However, this may not be the case if the camera uses interpolation in creating the image, this will create larger file sizes. To put it simply, interpolation invents pixels were none existed and inserts them in between the originals to increase the image s size. This invention of new data can be done quite smoothly, establishing a subtle gradation of values between the optical data, using interpolated data. Or, it can introduce noise, artifacts and other imperfections that actually degrade the image. When selecting a digital camera or any other device that uses an image sensor, always look for the optical resolution, not the interpolated resolution. Megapixel is the subjective term for the clarity of an image. It generally applies to a sensor and commonly refers to the resolving power achieved. A 6 MP (megapixels) SLR is superior in its resolving power to most 35mm 200 and 400 color print films, and is very impressive against some very good slide films. But a correctly used 35mm SLR, loaded with Fuji Velvia film and using a very good lens, will outperform a Canon EOS D60 (a 6.3 MP camera). Move the argument up to the larger film formats, such as 6x6 and 6x7 and the many sheet film formats, and digital is lagging far behind, even with the outstanding Canon EOS 1Ds. Properly used Velvia with a Hasselblad (medium format camera) using a good Zeiss lens will be better than 11 million pixels. Some people argue that a 35 mm frame has 16 million pixels (or 16 Megapixels). This argument is based on pure mathematics and is argued to be true. However, the basis from which the figures are worked out is debatable. Where do those figures come from? When lens makers and optical labs test lenses for their resolving power, they use a very slow technical film that has extremely high contrast and is almost solely used for testing lenses; a good example is Kodak Technical Pan, which can be used in the field if you are extremely careful to develop it slowly and at a high dilution of developer in order to stop it from New Technologies Graphic Communication 302 Winter Quarter, 2003 Page 55

8 turning into pure black and pure white. The images usually taken in these lens tests are of black and white lens test charts which have ever narrowing lines that allow the lab to determine how many line pairs per millimeter a given lens can resolve. The contrast of these test charts is 1:1000 as the only shades are solid black and solid white; precisely the reason that technical films are so contrasty in the first place. Other films can be tested in these conditions, Fuji Velvia for example, reputedly scores more than 160 lp/mm in such tests. These tests are the origin of the figures quoted and the reason that 35mm film being capable of 16 MP is a little questionable. Let s move the test into the real world where contrast rarely gets above 1:1.6 and where many of us shoot handheld and using relatively low cost SLR lenses. The situation alters dramatically, and in these conditions, Velvia s resolution drops dramatically to around the lp/mm mark. Using a tripod and stopping the lens down to its optimum aperture will get another lp/mm, but only a high quality lens, a solid tripod, a shutter release and a mirror lockup will get you close to lp/mm. This is off of the lab tests, and so it is also off of the 16 MP of image info. The Canon EOS D60 can resolve 67.7 lp/mm in a 1:1.6 contrast scenario in optimum conditions, not far off the best that Velvia can do, though handholding, using cheap lenses, and raising the ISO setting can hobble it back down to lp/mm, just as with 35mm. In the real world of color, 10 million pixels is as good as 35mm film, better in some cases, not as good in some others, but mostly good enough for quality photography. A 10 MP image has equivalent sharpness, if not slightly better than, good 35mm slide film. With the new Kodak DCS Pro 14n, a 14 MP, full-frame CMOS sensor, 35mm film is beat in sharpness, but with 36-bit color, it still doesn t reach the same tonality. If you ask people what is the most important specification of a digital camera in terms of image quality, most will tell you megapixels. That s correct, but it doesn t paint the whole picture. Instead of thinking megapixels, you should look at the maximum image size. Why the maximum size? To get the highest quality printed image from your camera, you must shoot at the maximum size. Most digital cameras capture information around 72 pixels per inch (PPI) regardless of their total pixel count and overall resolution. 72 PPI defines digital photography and yes it is both with still video cameras and digital camcorders. Computer monitors for the most part also playback at about 72 PPI. The total number of pixels used in the X and Y direction measures camera resolution. If each camera captures at 72 PPI and each camera captured exactly the same image, then the ones with more pixels would resolve more information. Since they display their images at 72 PPI on the computer screen, the ones with higher resolution will display at a Page 56 New Technologies Graphic Communication 302 Winter Quarter, 2003

9 larger size on a monitor. You can always reduce a picture in size using software, but you cannot enlarge a digital image larger than its original size. If you do you will lose quality because it is already at its maximum resolution. To produce a 4 by 6 inch print of photographic quality, a digital camera would require 2.1 Mega Pixel resolution (standard photographic quality print resolution is 300dpi). This is well within the capability of current digital cameras in the consumer market. Rendition of tone and enlarging capabilities depend on the limiting resolution. With film, the limit is grain size after processing plus loss of detail in enlargements. With digital cameras, it is defined by physical pixel size on the sensor and noise level. CCD Size Maximum Image Resolution High Quality Print (300 ppi) Good Quality Print (240 ppi) 1.3 Megapixel 960 x x 4.27 (8.1cm. x 10.9cm.) 4 x 5.3 (10.2cm. x 13.5cm.) 2.11 Megapixel 1200 x x 5.33 (10.2cm. x 13.5cm.) 5 x 6.67 (12.7cm. x 16.9cm.) 3.34 Megapixel 1536 x x 6.83 (13cm. x 17.3cm.) 6.4 x 8.53 (16.3cm. x 21.7cm.) 4.1 Megapixel 1704 x x 7.57 (14.4cm. x 19.2cm.) 7.1 x 9.47 (18cm. x 24cm.) 5.2 Megapixel 1920 x x 8.53 (16.3cm. x 21.7cm.) 8 x (20.3cm. x 27.1cm.) 11.1 Megapixel 2704 x x13.55 (22.9cm. x 34.4cm.) x (28.6cm. x 43cm.) Megapixel 3000 x x 15 (25.4cm. x 38.1cm.) 12.5 x (31.8cm. x 47.6cm.) As we stand right now, professional digital cameras have achieved, and in some cases surpassed, the sharpness of 35mm film cameras. Digital cameras have the advantage in dynamic range also, by almost a factor of two. The only thing holding back digital from completely replacing film is image tonality and image clarity. The improvement in tonality is limited for digital cameras, as they can never truly equal the infinite tonality that film can possess. Once digital camera manufacturers figure out how to reduce noise produced by the sensors, then digital cameras will have achieved 35mm film performance. So digital cameras are one step behind 35mm. This chart shows typical CCD pixel dimensions, resolutions, and the maximum size of images derived from these sensors. The graphic arts industry standard of 300 ppi for print is used as the definition of High Quality, while a lower-resolution option is presented. This lower-resolution option may be appropriate to ink-jet printing. New Technologies Graphic Communication 302 Winter Quarter, 2003 Page 57

10 Once 35mm is surpassed, it seems that medium and large format will not be far behind because the technology is within our grasp. Sources 1. Amen, Saeed. Digital vs. Film Cameras. Posted December 23, Retrieved February 4, 2003, from 2. Canon U.S.A., Inc Clark, Roger N. Image Detail (How Much Detail Can You Capture And Scan). Posted April 8, Retrieved February 3, 2003, from 4. Cook, Jeffrey A. Kodak Technical Sales Representative (personal communication, February 18, 2003). 5. Eastman Kodak Company Fuji Photo Film U.S.A., Inc Geoffrion, JR. Understanding Digital camera Resolution. Posted October Retrieved February 3, 2003, from 8. Goldwyn, Craig. Resolution and Print Size. Posted October Retrieved February 10, 2003, from 9. Grotta, Sally Wiener. Digital vs. Film: The Real Down-Low. Posted June 25, Retrieved February 4, 2003, from Hecker, Miles. Digital Camera Image Quality. Posted January Retrieved February 3, 2003, from Kimbler, Del. Digital or Film: Focus on Fundamentals. Posted December 30, Retrieved February 4, 2003, from Koren, Norman. Understanding Image Sharpness. Posted November 18, Retrieved February 3, 2003, from Lipe, Jessie. Foveon Public Relations Representative (personal communication, February 21, 2003). 14. Mainyu, Angra. Digital vs. Film, Why? Posted January 22, Retrieved February 4, 2003, from Ngo, Jim. Prints From Digital Cameras. Posted September 7, Retrieved February 4, 2003, from Nikon USA Photo Binbook. Digital Range vs. Film. Posted July 15, Retrieved February 4, 2003, from Reichmann, Michael. Pixel Count and Imaging Chips. Posted December Retrieved February 3, 2003, from Reichmann, Michael. Understanding Sharpness. Posted June Retrieved February 4, 2003, from Page 58 New Technologies Graphic Communication 302 Winter Quarter, 2003

11 20. Sony Corporation of America Unknown. Digital Photography vs. Traditional Photography. Posted June Retrieved February 3, 2003, from New Technologies Graphic Communication 302 Winter Quarter, 2003 Page 59