Worksheet for Benchmarking Assignment



Worksheet for Benchmarking AssignmentPlease complete this worksheet and submit it via the D2L Dropbox for the Benchmarking Assignment. Document filename: yourname_benchmarking.doc.Susan Edwards susanedwards_benchmarking.docNOTE: Where specific commands are required in the imaging programs, I’ve tried to supply directions in parentheses (e.g., Choose Image--> Mode --> Select Grayscale). These programs may vary slightly based on the program, version, and operating system. If you have problems, please see the Program Help. If you’re still confused, contact me and I’ll do my best to supply the appropriate commands.For questions 1-2, refer to image named “Wedding_File0003c.tif”Describe the original photograph. (Refer to the Cornell tutorial for appropriate terminology)Based on what you see (defects and all), what kind of original was this likely to be? The original is a continuous-tone monochromatic image—more specifically, a photograph. Most of the edges of the elements in the image are defined by soft, and subtly varying tones. It appears to be printed on a cream-colored paper. The pulp-textured surface of paper is visible in two areas: along the edges of the photograph where there seems to be some damage to the surface, and in the darkest areas. In addition, at the bottom of the image, the edge of the photograph itself is visible. This may be a gelatin silver print, or a print created by another photographic process printed on paper.What are the important characteristics of this type of original? (e.g. details, exposure, color cast, photographic finish, dynamic range, etc.) How would you characterize it tonally?Photographs in general can have fine details, a broad dynamic range, and many tones represented. A photographic negative can capture a lot of detail, but an under- or over-exposed print from the negative can obscure detail. This print appears to be properly exposed, since details are visible in both the light and dark areas. It may be possible to bring out more details from the dark areas by brightening the image, however, increasing the brightness of the image in PhotoShop results in a loss of detail in the white areas, especially in the embroidery below the neckline of the bride’s dress. Photographs can also display areas of blurring, where details are not captured, depending on the exposure, and movement of the subject. Blurring is visible in this image in the bouquet held by the bride, especially in the lower branches hanging in front of her skirt.Tonally, like most photographs, this image displays broad dynamic range—the white is very white, and the darkest dark areas are completely black. There are also many tones represented within that range, resulting in high tonal density. There is a lot of detail visible in both the lightest and the darkest areas of the image. The lace details in the bride’s white veil are visible, as well as the crinkled texture of her gown, and the wrinkles on both of the subjects’ foreheads and around their eyes are discernable. In the darkest areas, individual hairs on the bride’s head are almost discernable, as are wrinkles on the groom’s coat, and the pleats in the curtains behind the couple. The image is thus neither high key, nor low key, but displays a balanced tone.In addition to the details visible in the image captured in this photograph, details of the photographic object itself may be worth capturing. These include the obvious white-bluish colorcast visible around the edges of the print, probably caused by physical damage or deterioration over time. The specs of lint and dust on the surface, the brown spots visible on the groom’s teeth, and the fingerprint on the surface of the paper (visible at the upper right edge), are also details of the object that may be valuable information to capture in a digital scan.We can see that, although the photograph is a monochromatic image, there is color discernable in the photograph itself. The cream-colored paper shows through in the light areas, and the damage visible on the edges of the photograph has a bluish cast, which is also continuous and highly varied. In addition to this bluish colorcast, there are several small imperfections that display color, like the brown marks on the groom’s teeth.The image seems to be printed on matte paper, as evidenced by the visible grain of the paper and lack of reflection on the surface. If you were to digitize the original image, what characteristics would be most important to capture and how would that relate to the choice of resolution and bit-depth?The most important characteristics of this image to capture are the density of tone and broad dynamic range, and the details visible in the dark and light areas of the image, and on the photographic print itself. Capturing the color is also important here. From a technical-photographic standpoint, the color of this photograph is significant. The color provides crucial information for dating the image, determining the chemical process that was used to create it, and understanding the history of the photograph’s deterioration. However, scanning in color at high bit-depth, as opposed to 8-bit greyscale, would increase file size. So, depending on how the image will be used and for what purposes, and what kind of resources are available for scanning and for storage of images like this, I might decide to forfeit high bit-depth color capture. Thus, given the significance of the color, and if the image is being digitized for archival purposes, I would choose color scanning at 24-bit, since the many tones of the photograph, and the subtlety of the colors are both best captured with a high bit depth. Because the image has colors and a lot of detail, I would also want to use high resolution in order to capture the smallest details, at 400dpi or higher.If the dimensions of the original are 3” w x 5” h:Calculate the pixel dimensions it would take to scan the document at 300dpi.900 X 1500 pixelsCalculate the pixel dimensions it would take to scan the document at 600dpi.1800 X 3500 pixelsWhat approximate size, in both bytes and kilobytes (remember to use a factor of 1024, not 1000 when converting from bytes to kilobytes), would the file be if you scanned it at:Refer to the formulas in the Cornell tutorial: basic terminology> key concepts> pixel dimensions> file sizeNOTE: Include the formula you used, and your calculationsImage Type300 dpi (in bytes) 300 dpi (in KB)600 dpi(in bytes) 600 dpi (in KB)1-bit bitonal168,750 bytes165 KB675,000 bytes659 KB8-bit greyscale or color1,350,000 bytes1,318 KB5,400,000 bytes5,273 KB24-bit color4,050,000 bytes3,955 KB16,200,000 bytes15,820 KBFormula = (h x w x bit depth x dpi2) / 81-bit bitonal 300 dpi 3 x 5 x 1 x 90,000 = 1,350,000 1,350,000 / 8 = 168,750 bytes168,750 / 1024 = 164.7949 KB600 dpi 3 x 5 x 1 x 360,000 = 5,400,000 5,400,000 / 8 = 675,000 bytes675,000 / 1024 = 659.1797 KB8-bit greyscale/color300 dpi 3 x 5 x 8 x 90,000 = 10,800,00010,800,000 / 8 = 1,350,000 bytes1,350,000 / 1024 = 1318.3594 KB600 dpi 3 x 5 x 8 x 360,000 = 43,200,00043,200,000 / 8 = 5,400,000 bytes5,400,000 / 1024 = 5273.4375 KB24-bit color300 dpi 3 x 5 x 24 x 90,000 = 32,400,00032400000 / 8 = 4,050,000 bytes4,050,000 / 1024 = 3955.0781 KB600 dpi 3 x 5 x 24 x 360,000 = 129,600,000129,600,000 / 8 = 16,200,000 bytes16,200,000 / 1024 = 15,820.3125 KBOpen image: text_and_photo.jpg What are the important characteristics of this type of original? (e.g. color cast, condition, dynamic range, etc.) How would you characterize it tonally?This is a mixed-type document, which contains both printed text and a halftone image printed on paper, probably in a book. The printed text portion of the image has distinct edges, and is bitonal, containing only black and white tones. This kind of printed text can be relatively easy to scan because of the very consistent nature of the document’s tones. The halftone image on the page is monochromatic, but it displays a wider variation in tone than the text areas, including a range of greys in addition to black and white. Because of the grid of dots that make up the halftone image, it can be difficult to capture, as a moiré effect can be created when the grid of the halftone dots clashes with the grid of pixels in the digitized image.Both portions of this document display a high dynamic range, with very white whites, and very dark dark areas. While there is a range of tonal density in the halftone image on the page, the image does not display a very broad number of tones. The halftone image has little detail in the shadows, or in the white areas. There is little to no color in the document (other than black and white).The document appears to be in a very good state of preservation. The text and image are crisply printed, and the white medium (presumably paper) does not show any fading, or discoloration. The whites are evenly and consistently white throughout the image.The paper that the image and text are printed on appears to be thin, as there is a slight bleed-through of text from the other side of the page visible.If you were to digitize the original image, what characteristics would be most important to capture and how would that relate to the choice of resolution and bit-depth?The most important characteristics to capture in this image are the clarity and readability of textual information, and the detail and range of tones in the halftone image. We may want to ensure that the text is captured with enough clarity that Optical Character Recognition (OCR) software could transform the text into machine-readable and –searchable text with a high percentage of accuracy. In addition to potential to OCR the text, the presence of the halftone image will drive the choices of resolution and bit-depth, because halftone images are difficult to capture without causing a moiré effect.There doesn’t seem to be any significant color information in the document, so a color scan is not necessary. The halftone image requires greyscale scanning, as would any OCR performed on the text. With greyscale scans, we have a choice of 8- or 16-bit depth. Because the range of tones in the halftone image is not great, 8-bit scanning should suffice. The resolution would also be driven by the halftone image and the OCR of the text. Both of these factors indicate using a high resolution of about 400dpi. How would you measure those characteristics, so that you could benchmark for them?In order to determine the resolution that would avoid creating a moiré effect on the halftone images, we would use a halftone screen finder to measure the screen ruling on the halftone. It is suggested to scan at four times the screen ruling of the halftone. In order to measure the resolution of the text portion of the document, we would measure the height of the smallest character in the text—this is typically the letter ‘e’—and then use that measurement to determine dpi required to attain a high Quality Index (QI). We could use an online calculator, such as the Image Quality Calculator () to determine optimal resolution to capture the text.In order to measure the range of tones in the image, we would use a historgram in a scanning device or imaging software. Or, a tonal range target could be used to visually assess the range of tones in the document.Open image: Proust5folder1page1.jpgWhat are the important characteristics of this type of original? (e.g. color cast, condition, dynamic range, etc.) How would you characterize it tonally?This document is a manuscript page, with text handwritten in ink. The edges of the handwritten text are soft – they do not display the sharpness of a machine-printed text. There is also tonal variation within the lines of the ink handwriting, due to a lack of consistency in the application of the ink to the paper due to fading over time. The dynamic range of the image is broad, but not as broad as printed text or photographs. The dynamic range is characterized by dark areas that are not completely black, and light areas that are not completely white. Despite this, the contrast between the text and the page is high. While the number of tones in this document is not as great as those seen in a photograph, there are a lot of tones represented within the narrow ranges of the paper and the ink.Manuscript documents like this can also display significant color information. In this case, the color of the paper, and the ink lines of handwritten text can provide information about the creation and age of the document. Color can also give clues to the condition and history of the document, by showing areas of foxing, staining, and aging of the paper, and color fading of the ink used.This document also has areas of wrinkling where the paper has been folded. There is a sharp vertical crease on the left edge of the document, as well as more subtle wrinkling horizontally across the middle of the page, apparently from being folded in half. If you were to digitize the original image, what characteristics would be most important to capture and how would that relate to the choice of resolution and bit-depth?Detail and clarity of the handwritten text would be the most important characteristic to capture in this manuscript page. Manuscript pages can be scanned with bitonal, greyscale, or color capture. In this case, bitonal capture would not be a good choice because of the imperfections in the paper—staining and wrinkling—would probably create artifacts on the page and make part of the text unreadable. Thus, scanning in either greyscale or color would require a bit depth of at least 8-bit. Because this document does not display a great deal of tonal variation, 8-bit capture may be sufficient even for color capture. In order to capture the detail and clarity of the handwritten text, a high resolution is recommended.The second most important characteristic of this document to capture would be the color. Color capture would provide significant information about the history of this document, its preservation status, and its age. If choosing to capture color, the bit depth could be raised from 8-bit to 16- or 24-bit. However, this may not be necessary as the range of colors in the document is not great. Raising the bit-depth would result in a filesize increase. How would you measure those characteristics, so that you could benchmark for them?To measure the detail in the handwritten text, I would measure the width of the finest stroke, and use the greyscale/color formula to achieve a high QI. Or, I would use the 2px rule and make sure that the finest detail would be covered by 2 pixels.In order to measure the range of tones in the image, we would use a historgram in a scanning device or imaging software. Or, I could use a tonal range target to visually assess the range of tones in the document.Open images: “Color_target01_rcm1087047322.jpg” and “Color_target02_rcm1087046753.gif” Why might a color/grayscale target be included with these images?Targets provide an objective measurement of color and tone values, and are used for several reasons, including equipment calibration and image processing.The targets may be included here in order to verify that the scanner is still in calibration. The colors and tones on the targets would produce known digital values in the scan, which could be verified in the imaging software. If the targets are not producing expected digital values, it would be an indication that the scanning equipment is out of calibration and needs to be adjusted. The target may also be used to calibrate an individual monitor when viewing the image.The targets may also be included with the individual images so that they can be used during image processing to perform digital corrections on the image. They also provide verification of the quality of the scan itself. Finally, the targets may also be used to benchmark the colors and tones in an image and in a larger collection of similar images. What is the bit depth of each image? (Image-->Mode)Both images are 8-bit 6) For image: “Color_target02_rcm1087046753.gif” Open the image histogram (Window--> Histogram). Experiment with the brightness/contrast adjustments (post-processing options). (Image-->Adjustments). What happens to the image when you adjust the brightness? (Be sure to look at the color bar too). Explain why brightness/contrast functions might be used and why they should be used sparingly.When I adjust the brightness in the image to increase brightness, I lose detail in the brightest areas of the print, and the whole image seems washed out. Similarly, when I adjust to decrease brightness, I lose detail in the darkest areas of the print. This is very visible on the tonal scale. Brightening the image to the maximum amount makes the last four lightest grey boxes on the right completely merge with one another as one white strip. Darkening to the other extreme makes the darkest boxes on the target, at the left, merge into one another as a black strip. I can also see changes in the histogram, as I move the slider to either extreme, from light to dark; the number of tones in the mid-tones decreases, and the tones at the extremes increase. Adjusting the contrast, moves all of the tones in the image either towards the center (less contrast), or out to both ends of the spectrum (more contrast).Adjusting the brightness and contrast can be used to bring out details in the light and dark areas of the image. However, bringing out detail on one end of the spectrum seems to mean losing detail on the other end. So achieving a balance is key. One should be aware of not sacrificing too much detail on one end of the spectrum in order to reveal detail on the other end.Now focus on the caption of the picture (View > Zoom In) Do you notice anything about the text? Describe what you see. Discuss some reasons why the text might appear as it does.The text is pixelated. There are two lines of text, and the smaller, second line below the title is so badly pixelated that it is unreadable. There are also compression artifacts visible around the text. The parameters chosen for this scan were probably optimized for capturing the image, at the expense of the text on the page. This would have been done because the image was perceived as more significant than the text. In addition, when the image was probably captured at a high resolution as a tiff, and has been compressed to a gif, resulting in loss of pixels. Open image “Beach_File0001c.tif”, note the image size. (This info is usually found in the lower left hand part of the Photoshop window)897255100965Each exercise below should be performed on this file020000Each exercise below should be performed on this fileUsing Photoshop or Photoshop Elements, duplicate the image (Choose Image--> Duplicate). On the copy, change it from a 24-bit color file , to 8-bit, to 1-bit, 400 dpi, B/W image via a 2-stage process:Choose Image--> Mode --> Select Grayscale. Discard color info? Choose Discard.Choose Image--> Mode --> Select Bitmap. A pop-up will appear. Make sure it says “400” in the output “pixels/inch” box. What happens to the file size?Step 1 (24-but color 8-bit greyscale) – it went from 16MB to 5.3MBStep 2 (8-bit greyscale -bit, 400dpi b/w) – it went from 5.3MB to 684KBStill on the copy, choose “Save for Web” (File-->Save for Web) or “Save As” (File-->Save As). If using Save for Web, Choose to Save it as “JPEG,” “Medium Quality” (or quality=30). Rename the file! Save it as “Beach_file0001_jpg.” What happens to the file size?Because I am unclear whether I am supposed to continue here with the 1-bit, 400dpi b/w image, or start again with a new copy of the tiff, I am doing both, and giving you both answers here:When I performed this operation on the 1-bit, 400dpi, b/w image from 2b step above, the filesize went from 684 KB to 2.4MB.When I performed this operation on the original Tiff, the filesize did not change at all. The Tiff was 16MB, and the new jpg is also 16MB.Open the original file (File0001c.tif) and the new image created above in 4b (file0001_jpg) and look at the images. [View them at about 16.7% - View--> Zoom in or out to get there.] Can you perceive a difference in quality in the two? Again – I’ll compare both the jpg created from the 1-bit file, and the jpg created from the original tiff:1. Original tiff compared to a jpg created from the 1-bit file from step 2b There is a big difference in quality between these two images. The original image has a yellowish colorcast, and the new image is greyscale. The original image has very smooth tones, and I can’t see any grain at all at 16.7%. The jpg has visible dots in the sky and on her thigh, and a hot spot on her forehead. Her skin looks splotchy in the new image. There seems to be a bit more contrast in the new image.2. Original tiff compared to jpg created by saving the original tiff for web at medium quality When viewed at 16.7%, there isn’t much difference in quality between these two images. There is a slight splotchiness visible in the sky of the jpg. Otherwise, there is little visible difference in quality between the two.Look at the histograms for the two images (Window-->Histogram). Click on one image and view the histogram. Now click on the other. Did you observe any differences? If so, what?Again – I’ll compare both the jpg created from the 1-bit file, and the jpg created from the original tiff:1. Original tiff compared to a jpg created from the 1-bit file from step 2b There is a big difference in the historgrams of these two images. The original historgram has a smooth curve—a ‘mountain’ rises towards the right with smooth sloping curve going down on the left side, and a slightly steeper curve on the right side. The historgram of the new image looks like a bar graph—it's a series of straight lines, with the lines on the right taller than those on the left. Basically, the original Tiff displays great density of tones, and the derivative jpg image has a great reduction in tonal density.2. Original tiff compared to jpg created by saving the original tiff for web at medium quality The histograms of these two images are similar, but the histogram of the jpeg does not show the same smoothness as that of the tiff. The jpg histogram shows the same general curve as the tiff’s histogram, but the edges of the curve are rough, and spikey. Presumably, the missing areas of the histogram curve in the jpg correlate with the pixels that have been tossed away in compression.Repeat step a. Change the image to grayscale (Choose Image--> Mode --> Select Grayscale). On the new copy, choose “Save for Web” and this time, save it as a GIF file (rename it Beach_ file0001_gif).Open all three files onto the Photoshop workspace. Compare all three. Are there compression artifacts visible in the JPEG version? In the GIF version? Make sure to provide details to support your answer.533400109855NOTE: include samples from your images here to support your answer020000NOTE: include samples from your images here to support your answerComparing all three images, it’s pretty clear that there is a huge loss of detail in the jpg created from the 1-bit derivative (Figure 1).Figure 1 - Comparison of original Tiff (left), gif derivative (middle), and a jpg created from a 1-bit black-and-white derivative (right)Comparing the original Tiff to the gif, I don’t see any compression artifacts. Figure 2 shows a comparison of details the girl’s face at 100%. The images seem comparable in representation of details and tonal range. The gif seems to show a little more graininess, which is visible in the sky behind her head, although it is hard to tell if this is present in both images, and more visible in the gif because the contrast is more distinct in the greyscale image than in the color of the Tiff.Figure 2 - Comparison of original Tiff (left) to gif derivative (right)Even when zooming in to 300% (Figure 3), there is no discernable difference between the Tiff and the gif. The level of detail is comparable, and even defects in the original Tiff scan are reproduced in the gif.Figure 3 - comparing detail of original Tiff (left) to gif (right) with both zoomed in at 300%When comparing the original Tiff to the jpg created in Step 2b (by creating a 1-bit black-and-white version of the Tiff then compressing to jpg), there is a dramatic difference. The Jpg is very grainy, with little detail. As seen in Figure 2, where the hairs on the girl’s head are individually discernable in the Tiff, they become a blur of grey in the jpg. There is clearly a loss of detail and pixels, which is very evident in a flaw in the photograph that all but disappears in the jpg. In Figure 4, this flaw is visible in the Tiff to the right of the girl’s head—it is a circular, greyish smudge located almost directly below the black mark in the photo. In the jpg, this smudge has almost disappeared.Figure 4 - Comparison of original Tiff (left) with a Jpg created from a 1-bit black-and-white derivative.The graininess and lack of detail of the jpg becomes extreme when we zoom in to 200% and compare it to the original Tiff, as seen in Figure 5. In this comparison, compression artifacts are visible in the Jpg along the bottom of the detail, below the ladder, and along the bottom rung of the ladder.Figure 5 - Comparison of details viewed at 200% from original Tiff (left) and Jpg created from a 1-bit black-and-white derivative (right)When comparing the original Tiff to a jpg created directly from the original tiff, the differences are harder to see, although some compression artifacts are definitely visible. In Figure 6, compression artifacts are visible in the shoulder area, just between the girl’s armpit and the edge of her bathing suit (this is hard to see in the screenshot here – it was more visible in Photoshop!).Figure 6 - comparison of detail from original Tiff (left) and a Jpeg created from that Tiff at medium quality (right)When viewed at 300%, the compression artifacts in the Jpg become more apparent. In Figure 7, they are visible around the edges of the girl’s chin, and the edges of the shadow on her neck.Figure 7 - details seen at 300% of original tiff (left) and derivative Jpg (right)Choose a particular section of the image. Using the Zoom tool (or View-->Zoom In), zoom into that area of both pictures. What percentage must you zoom to before seeing clear pixilation in Beach_file0001c.tif? In Beach_ file0001_jpg.jpg? In Beach_file0001_gif.gif?For the Tiff and the gif, I have to zoom in to 200% before I can see clear pixelization.For the jpg created directly from the Tif, I start to see pixelization at about 185%.For the jpg created from a 1-bit derivative of the Tif, I don’t see pixels until 200%.Briefly give your view on the trade-offs between file size and quality of images. File size seems to correlate directly to the amount of detail captured in an image. Bit-depth, resolution, and pixel dimensions all affect the amount of detail captured in an image. It is important to consider the type of document being captured, as well as the purpose of the digitization, and available resources and storage, when deciding on the optimal file size.A high-quality, high-resolution Tiff is desirable for archival purposes in order to capture as much detail as possible for future uses and technologies, which are unknown today. The vast amount of detail captured in a high-quality, high-resolution Tiff (which thus has a large file size) is only discernable with the human eye when zooming in very closely. So, for images derived from the archival master for use today, the file size should be matched to the size of image that is needed for any given display. At small display-size, reducing the file size by saving as a gif, will not result in much appreciable loss in details. ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download