Image Enhancement Using Adobe Photoshop 2.5

Karen Holmes

LIS 296A
Howard Besser, Instructor
Final Project
4/29/94

NOTICE

GENERAL INFORMATION ABOUT THE PHOTOSHOP PROJECT

INTRODUCTION

My work on this project began when Professor Howard Besser suggested I contact Mr. Barclay Ogden, Head of Conservation, to discuss simulated restoration of photographic material using digital technology. A brief meeting with Mr. Ogden led to some interesting questions about the use of such technology. He was very curious to know if simulated restoration actually "restored" an image, or if in reality it created something false, or new in the image's place. He also liked the idea of having an undamaged original - then damaging it, scanning it, and finally attempting to get it back as closely as possible to its original condition using digital technology. Having an undamaged original would provide a "control" that is usually absent in instances of actual restoration. Seldom does the restorer know what an image looked like before it sustained damage.

Several weeks after my initial discussion with Barclay, a larger and more thorough meeting took place in the Conservation offices. Present were Barclay and myself, as well as Nancy Harris, Conservator of Paper and Photographs, Dan Johnston, Photographer in the Photographic Services Department, and Ann Swartzel, of the Conservation Department. Together we brainstormed on possible types of damage and images to explore. As we aimed to have roughly a dozen restoration problems to choose from, we came up with the following possibilities: first we agreed it would be interesting to have a relatively sharply focused and detailed subject, as well as a softer, smoother and more gradually tonal subject. To these two types of photos we could apply the damage of fading, surface abrasion (both to print and negative), dust, and missing emulsion.

It was arranged that Dan would supply the prints from his own collection in order to avoid any possible copyright problems that might develop if the images were ever to be placed online for public access. Nancy agreed to attempt the creation of missing emulsion, while I agreed to try my hand at print abrasion. Dan worked on the problems of dust, fading and negative abrasion.

While prints and damage were being made, I began learning Adobe Photoshop 2.5 - primarily through its Tutorial. I also met with Susan Stone at the Museum Informatics Project to discuss the possibility of using the scanner and printer there. It was at MIP that I began to be aware of two major obstacles on the path to a smooth project - space and time.

SPACE

Susan had been involved in some scanning projects before, and had an idea that each image would take at least 10MG of storage space. This created a problem for me, since I had hoped to store images on floppy disks in order to take them home and work on them there. Susan suggested finding out how much storage space I could get at the Library School, and also running some scanning tests to see how much room I might actually need.

A few days later, we proceeded with some printing and scanning tests. Both Susan and Natalie Munn, a Graduate Student Researcher at MIP, urged me to work at a resolution of at least 600dpi in order to have sufficient pixel quality with which to work. A scan of an 8x10 black and white image at 600dpi resulted in a file of over 25MG. In relaying this horrifying information to Professor Besser, he suggested scanning full frame images for context (both before and after restoration) at a lower resolution. He also mentioned working at home on higher resolution small details that would fit on a floppy diskette, and then transferring techniques that worked well to larger files stored at MIP or the Library School. In the meantime, Roberta Epstein at the Library School said she could give me about 30MG of storage space, and Susan and Glen at MIP said they could provide a similar amount. At approximately 6MG for an 8x10 scanned at 300dpi, plus approximately 9MG for medium-sized details scanned at 600dpi, I quickly realized that 10-12 images in all their various permutations were quickly going to eat up my allotted storage space. I began looking into backup tape possibilities and compression. By now I had finished the Tutorial, and the prints from Dan and Nancy were ready. It was time to start the actual scanning, restoration and printing. This led to the second obstacle of time.

TIME

After some experimentation, it became apparent that it took approximately 5 minutes each time I scanned an each image, no matter what its size, an average of 2-5 hours to do the restoration work, 5 minutes to print 1.3MG of 600dpi information, 20 - 30 minutes to print 6MG of 300dpi information, and about an hour to print 25MG of 600dpi information. This meant I could expect to spend at least 4 hours of time on each image. Add to this the fact that MIP, the Library School and I were only compatible at certain hours and I began to make some decisions about how many images to work on, and how many digital prints to make.

DECISIONS

I decided that I would aim to explore 6 images. I asked Barclay which types of damage he was most interested in, and he felt that fading and abrasion were the most common restoration problems and should perhaps be the ones I dealt with. I chose one each of detailed and tonal fading photos, as well as one abraded tonal photo. I had already started work on a photo with dust, so I decided to also pursue that in both a detailed and tonal subject. Finally, I was intrigued with Nancy's missing emulsion photo which dealt with text, hitherto unseen in my examples. This formed my sixth image.

As far as providing a low resolution full frame image both before and after restoration, I decided to print the before version only. I could have only shown the restored area in context by saving a damaged original scan at full frame, and then pasting a restored detail on top of it. However, this would have resulted in another 6MG of storage for each image, plus at least another half hour of preparation and printing time for each photo. I feel that in all cases, it is clear where the details would fit in their original works without literally showing them.

Finally, I found that because of my inexperience with Photoshop, it initially took me so long to do the actual restoration work (about 6 hours for the tiny detail of dust on the sharply focused image) that I just did not have the time to repeat my corrections on a larger detail later. This also meant that I could dispose of some medium-sized details that I had stored at MIP that were taking up a good amount of space. Though I wish my restored details encompassed a larger area, I think that one can get a good idea of results. Also, my techniques would not have changed simply because I had more image to work on.

Because I did at least half of my exploration at home using floppy disks, I could coordinate my use of MIP's printer and scanner pretty well. I would first scan the images I needed to take home, do the restoration work, return to MIP to print my work out, and begin the process again. In the meantime, Glen (whose last name I am ashamed to say, I never got) moved my folder off of MIP's Quadra hard drive and onto a server. This gave me quite a bit of storage space and lessened my need to get things on tape, use compression, or use the space set aside for me at the Library School. It never became clear if the tape drives of the Library School and MIP were compatible, but I'd venture to say that they are not. At the end of the project, after throwing out as much as I could, I still had about 100MG worth of files on MIP's server. Though Glen has kindly allowed me to leave them there for the time being, I will eventually have to store them in some other capacity.

THINGS I WISH I'D DONE DIFFERENTLY

I was a little disappointed in the dark quality of the digital printouts. In fact, I urge anyone looking at the printouts to do so under a strong light in order to see detail. I think if I had spent some time experimenting with scanner settings, I might have been able to lighten them up a bit. Also, I think I spent too much time correcting things at the beginning that you would never see anyway. Other things like worrying about storage space, choosing the wrong tools, and working too slowly were a product of not having worked with Photoshop before. I think my skills definitely improved as I went along.

BACK TO THE CONSERVATION DEPARTMENT'S QUESTIONS

I would have to say that it would take a very skillful person working with Photoshop to not falsify a digital image in some way in the process of restoring it. It might be in a way that would never be noticed by the average person looking at an unfamiliar image, but especially in cases where it is not possible to look at an undamaged original, I think it is very easy to slightly change a wisp of hair, or an architectural detail. Though, again, these slight changes might not be noticeable to anyone but the staunchest critic. In any case, I do think that a person doing this type of restoration work has to have high standards and ethics, and must fight the temptation to give in to one's own artistic inclinations.

For example, what if it made better compositional sense to move an object over just a touch, or to eliminate a distracting and informationally negligible bit of someone's head? If only a shadow could be lightened, or a "hot spot" dimmed. One must think carefully about the results of giving in to such musings, and know from the start what it is that one is trying to achieve. If the end goal is to have a digitized image that matches the original as closely as possible, then "improvements" of composition must certainly be avoided. But what of contrast and brightness? If an original print has areas that are excessively bright, would it be true to the photographer's original work to change that? One must also consider the purpose of making particular digitized images accessible in the first place. If it is solely to act as a visual guide to the types of images in a collection, then perhaps it is permissible to dull a bright corner. I think many questions such as these need to be asked, and the answers will vary on a case by case basis.

Barclay was also interested in finding out if certain problems could be solved in a standardized way. That is, every time there is a dust spot, can it be eliminated in the same manner technically? I think that some general techniques apply consistently, but again, every photo must be taken on an individual basis. Tones, focus, detail, and damage all add up to create unique situations to be addressed.

A particular concern of Dan's had been that tonal relationships and detail both be maintained in the digital restoration process. In general, though I would like to see a better quality output to demonstrate this, tonal relationships and detail do show up quite well on the computer monitor - although details are slightly softened by having been broken into pixels.

THANKS

There are still a great number of issues to explore in image enhancement, but I feel that I made a good start and learned quite a bit. I couldn't have proceeded as smoothly as I did without the help of a great number of people. I would like to thank Howard Besser for suggesting this project and for getting me in touch with Barclay Ogden. I would like to thank Barclay for his swift setting up of intriguing and thought-provoking meetings and unflagging encouragement. Thanks also to Ann Swartzel for her enthusiasm, Dan Johnston for his beautiful prints and generosity of time, and to Nancy Harris for her challenging emulsion removal. I am also grateful to Lester Weiss at the Pacific Film Archive for his kindness, Photoshop advice and time, and to Roberta Epstein at the School of Library and Information Studies for her technical help. Finally, my extreme gratitude to Susan Stone, Natalie Munn and Glen at the Museum Informatics Project for their patience, use of equipment, materials, software and massive amounts of storage space. They made my work extremely enjoyable.

TECHNICAL SPECIFICATIONS FOR MATERIALS AND EQUIPMENT USED

COMPUTERS: Macintosh IIsi and Macintosh Quadra 950

SOFTWARE: Adobe Photoshop 2.5

PRINTER: Tektronix Phaser IISDX Dye Sublimation

(Primarily loaded with black and white transfer rolls. One trial printing was run using a 4-color transfer roll. Though the color version is in some ways more appealing, color transfer rolls are more expensive to use than black and white. Black and white was settled on for economic reasons.)

Though several tests were carried out with various printing settings, no appreciable difference could be ascertained in quality of output. As a result, Default Screens were specified (rather than Accurate Screens with or without default Frequency, Angle and Shape specifications). The Color/Gray Scale option was chosen rather than Black and White.

SCANNER: Agfa Arcus Professional Image Scanner. Settings used were:

Mode: Gray-Scale Original: Reflective Input: 300 ppi (or 600 ppi) Scale To: 100% Tone Curve: None Auto Density on Preview Size: Max. Area Resolution was not changed within the Photoshop Application. Cropping was primarily done in the scanning process rather than in Photoshop.

PHOTO-SPECIFIC INFORMATION

NAME OF PHOTO: Kids/Abrasion

PROBLEM(S) TO BE CORRECTED: Remove scratched areas. Attempt to maintain smooth gradations of tone in skin.

CORRECTIVE MEASURES TRIED WITH VARYING DEGREES OF SUCCESS:

ACTION: I began with the rubber stamp tool to "clone" areas of skin and hair. I would click on an unaffected area to store that as the sample I wished to reproduce over a scratched area. (I did this using the "aligned" option, which meant that my sample would move along parallel to the area I was replacing. If I had chosen the "non-aligned" option, my sample would be continually taken from the initial area I had selected.) Next I would click and drag through damaged areas, laying cloned areas down as I went. I usually worked with a medium-small and slightly soft "brush".

RESULT: Because this was my first attempt at cloning skin, I did not realize that dragging the rubber stamp tool tended to create a smudgy effect. As I went along, I saved my work frequently. This meant that I had more and more smudgy areas, and less pure skin tone from which to clone successfully. As I learned later, it would have been better to click frequently, giving "puffs" of cloned skin, rather than long smears of it. In succeeding photos with damaged smooth tone areas, these "puffs" in conjunction with short drags of the rubber stamp tool eventually seemed to work the best.

ACTION: For small, or very detailed areas of damage - I used fine, slightly soft brushes. Using the option key, I could toggle back and forth between brush and the eyedropper tool. With a click of the mouse, the eyedropper tool would store the color of the area I had just clicked. This would then be the color used when I toggled back to the brush tool and started painting with it.

RESULT: This seemed to be a good and useful combination of tools. The main thing to watch was that brush sizes and hardnesses weren't too big or fuzzy for the area being repaired (or that they weren't so small as to be ineffectual). Very large areas that were simply painted, rather than cloned, tended to look flat. If I didn't use the "undo" command quickly enough, I would sometimes have to go back over these flat painted areas with some cloned texture. Because the scratches in this photo were similar in their shape and direction, and very close together, I did actually create a custom brush shape for this particular job. I made a narrow, soft oval-shaped brush to follow the shape of the scratches. In retrospect, I am not sure that this was necessary. In fact, a larger burst of cloned texture would have served to cover more scratched areas at once.

ACTION: When repaired areas looked a little too sharp, I would try a variety of "blurring" filters. Selecting the place in question with the lasso tool, I would try "blur," "blur more," or "gaussian blur" from the Image menu.

RESULT: Usually one of the above filters would do the trick - "blur" being the least noticeable, and "gaussian blur" being the most blurry of the filters. Sometimes I would apply two " blurs" in a row, or one "blur" plus a "blur more." Different combinations worked in different areas.

ACTION: Conversely, I found the upper lip of the boy to be too out-of-focus after I had worked on the area immediately above it. Though I could have tried a sharpening tool or filter, my corrections had rendered the difference between lip and upper lip too indistinct. To correct this, I applied tiny dots of color along the edge of the lip with a very small, slightly soft brush.

RESULT: This technique helped redefine the edge of the lip, but it now looked too sharp. I corrected this by applying a blur.

ACTION: In some instances, repaired areas would seem too dark or light. I would then try the dodging tool (to lighten) or the burning tool (to darken). By hitting the option key, I could toggle back and forth between the two tools.

RESULT: Usually the use of these tools created an improvement. However, it was easy to go too far in either direction. These tools seemed to work best with a light hand.

MOST SUCCESSFUL (OR ULTIMATELY APPLIED) CORRECTIVE MEASURE(S): Rubber stamp tool to clone areas; fine brushwork for details; selected combinations of blurring, darkening and lightening.

GENERAL OBSERVATIONS: I was not really happy with my final results on this particular photo. The skin tones of the boy appear very blotchy and uneven. However, I got to the point where I felt I had already spent too much time on it, and needed to move on to something else. In retrospect, it does provide a nice contrast to my later, more skillful work, and shows that my technique did improve. If I had it to do over again, I would use the "puff" approach to cloning that I described above, in combination with short drags of the rubber stamp tool. Working with human faces on this project reiterated what I had already learned in actual drawing classes: it doesn't take much to change the shape of an eye, or the curve of a nose. When one's goal is to repair - and not to make something new - the subtle strokes that create these changes have to be watched very carefully.

PHOTO-SPECIFIC INFORMATION

NAME OF PHOTO: Kids/Dust/Hair

PROBLEM(S) TO BE CORRECTED: Remove hair and dust. Attempt to maintain smooth gradations of tone in skin, and maintain detail in shirt.

CORRECTIVE MEASURES TRIED WITH VARYING DEGREES OF SUCCESS:

ACTION: I began with the rubber stamp tool to clone areas of skin using the "aligned" option. I again chose the aligned option since I was trying to create subtle changes in skin tone, and wanted my sampling to follow along with these changes as they occurred. Since the damage was more localized here than in the "Kids/Abrasion" photo - I chose a large, soft brush which could cover the rather wide area of white created by the intrusion of the hair.

RESULT: As I had discovered at the very end of my work on the "Kids/Abrasion" photo, clicking the rubber stamp tool, rather than dragging it, created a more realistic texture of skin. In the face area, the only hazard was in choosing too large a brush for the rubber stamp to work with. In so doing, I reproduced the white glow of the hair as I followed alongside it while cloning. This was corrected by choosing a smaller brush. Other than this, the most difficulty I had was in the boy's shadowy neck area. Because the transformation from dark to light was subtle, and cut right in half by the white hair, I had a hard time avoiding a definite line between shadow and highlight right where the hair prevented me from seeing - and cloning - the actual changes in tone. I eventually worked around this by continually sampling and applying tone from either side of the white hair. A final smoothing out was achieved by changing the opacity setting for paint coverage to 50% (instead of the default 100% setting) in the Brushes dialog box. This allowed for a more gentle transition by providing a "wash" of tone instead of harsher flat areas of color.

ACTION: For the small, or very detailed areas of damage - especially in the shirt area - I used fine, slightly soft brushes. I did attempt some cloning, but in general the stripes of the shirt were so small and undulating, that the brush tool did a cleaner job of repair.

RESULT: By now, I had the hang of it. I didn't experience much difficulty in repairing the shirt.

ACTION: Just as for "Kids/Abrasion," when repaired areas looked a little too sharp, I would try a variety of the "blurring" filters. Selecting the place in question with the lasso tool, I would try "blur," "blur more," or "gaussian blur" from the Image menu.

RESULT: Usually one of the above filters would do the trick - "blur" being the least noticeable, and "gaussian blur" being the most blurry of the filters. Sometimes I would apply two " blurs" in a row, or one "blur" plus a "blur more." Different combinations worked in different areas.

ACTION: I again found that my work had created an indistinct area, this time between chin and neck. Using the sharpen filter from the Image menu just made the pixels clearer. After undoing "sharpen," I found that applying tiny dots of color along the edge of the chin with a very small, slightly soft brush helped separate head from neck.

RESULT: This technique helped redefine the edge of the chin, but it now looked too sharp. I corrected this by applying a blur.

MOST SUCCESSFUL (OR ULTIMATELY APPLIED) CORRECTIVE MEASURE(S): Rubber stamp tool to clone areas; fine brushwork for details, "washes" of cloned texture at 50% opacity for shadowy neck area; some selective blurring.

GENERAL OBSERVATIONS: I was much happier with the results of my second attempt at correcting damage in skin areas. The "puff" and short, light strokes of the rubber stamp tool seemed to work best here. I learned not to overwork an area and to avoid heavy-handedness. Right before you think you are done - stop.

PHOTO-SPECIFIC INFORMATION

NAME OF PHOTO: Kids/Variegated Fading

PROBLEM(S) TO BE CORRECTED: Adjust tones in over- and under-exposed areas of print so that bands of differing color are no longer discernible. Keep skin tones even.

NOTE: This was actually one of the first photos that I started working on. I would struggle with this one until I couldn't stand it, and then switch over to the "Kids/Abrasion" photo, and vice versa. By the time I sought outside help on how to correct exposure, I was starting to master some techniques that then led to a fairly quick cleaning up of remaining slight lines left between the bands of corrected light, medium and dark areas.

CORRECTIVE MEASURES TRIED WITH VARYING DEGREES OF SUCCESS:

ACTION: I began by trying to even out the tones of the entire image to match those shown in the middle of the print (the "medium band"). I consecutively selected the various bands of exposure using the marquis tool. For each band, I next selected Histogram from the Image menu.

RESULT: This gave information about average brightness, middle values, numbers of pixels, etc. for each area, but didn't seem to give a way to quickly adjust them. I was hoping I could duplicate information given for the medium band of exposure and apply it to the light and dark bands, but this didn't seem feasible in Histogram. That is, there was no place to simply plug in numbers.

ACTION: Next, using the marquis tool, I selected the medium-toned band. Then, from the Image menu I selected the submenu Map, and finally the submenu Equalize Entire Image Based on Area.

RESULT: The entire image bleached out. It almost looked posterized.

ACTION: I double-clicked on the marquis tool to select a feathering values of 8 pixels. I hoped that this would lessen the harsh line between bands of exposure (which it did to some degree upon application). I then selected the dark band with the marquis tool. Next I went to the Image menu, then chose Adjust, and then the Levels submenu. I adjusted the white and black points to even overall contrast. I then dragged the gamma (mid-tones) button to the left to lighten the selection, and bring it more in accordance with the tones of the medium band.

RESULT: This seemed to work fairly well. However, there was still a noticeable line between the dark and medium-colored bands where my selection hadn't been absolutely precise. When attempting to darken the light band of exposure, I followed the same procedure as just described. However, because there was so little tone in the light area, darkening seemed to add pixels that created a coarse-looking effect. To lessen this, I tried "blur" and "gaussian blur" from the submenu of the Filter menu - both before and after making corrections to remove the line between exposures. It seemed to work best to first make line corrections and then apply the gaussian blur.

ACTION: To attempt to remove the line between bands of exposure, I first tried both the dodging and burning tools.

RESULT: The effect of both tools was quite abrupt, and did not provide a smooth gradation of tone.

ACTION: Next I tried the smudge tool.

RESULT: The resulting smudging looked like it actually was smearing the pixels. The smudged areas looked very different from the untouched pixels next to them. The disparity was too great.

ACTION: Next I tried the rubber stamp tool with the non-aligned clone option. I set my opacity at 60% and tried a number of brush sizes and sharpness.

RESULT: This worked the best to blur the line between exposure bands. However, I felt there was still a lot of room for improvement, and didn't really know what to try next. I decided to call Lester Weiss, the Computer Expert at the Pacific Film Archive. Lester was kind enough to spend several hours with me on the Variegated Fading problem. After watching Lester play around with this, I proceeded as follows:

ACTION: Because the lines separating the varying exposures were fairly regular, I could get away with using the marquis tool for selection. However, in reality, the line between light and medium was not exactly vertical. I ended up selecting this using the lasso tool in both its freehand and straightedge mode. By holding down the option key and clicking the mouse, I defined the starting point of my straight line segments. While continuing to hold down the option key, I dragged along the edges, clicking to define points, and adjusting when necessary to make a less than vertical line. I also eventually selected irregularly shaped areas, such as the bright area above the left-hand boy's shoulder, for further brightness and contrast adjustments.

RESULT: I better fine-tuned my masked-off areas for exposure adjustment. I also used a feathering selection of 6 pixels which seemed to reduce and soften the line between corrected exposure areas.

ACTION: After selecting the areas on which I wished to work, I next I went to the submenu of Curves, under Adjustment in the Image menu. By clicking the Preview button, I could watch as adjustments were made to exposure on screen. Like the Levels dialog box - which I had tried earlier - Curves lets you adjust brightness, contrast, and gamma of an image. However, instead of making the adjustments using just three variables (highlights, shadows, and gamma), you can use the Curves dialog to adjust any point along the grayscale.

RESULT: By pulling and pushing on the "Curve" for the area I had selected, the tones lightened and darkened. When I thought they matched the unselected band of medium exposure next to them, I clicked o.k. Lester showed me how to save curves to use again, though I found that I was constantly readjusting anyway.

ACTION: Next, I adjusted some areas for brightness and contrast by choosing Adjust from the Image menu and Brightness/Contrast from the submenu.

RESULT: Not only did this bring brightness and contrast into line with the medium band of correct exposure, but it helped smooth coarseness in some cases. I found that as a result, I didn't have to use many blur filters.

ACTION: After correcting both light and dark exposures, I cleaned up the slight lines that remained between the former areas of dark, medium and light. I used a variety of cloning, blurring, sharpening and painting techniques already mastered in the preceding two "Kids" photos.

RESULT: The lines disappeared for the most part.

MOST SUCCESSFUL (OR ULTIMATELY APPLIED) CORRECTIVE MEASURE(S): Adjusted Curves, and Brightness/Contrast; rubber stamp to clone areas; fine brushwork for details; some selective blurring, sharpening, darkening and lightening.

GENERAL OBSERVATIONS: Though I had an initial hard time dealing with the problem of variegated fading, it actually turned out to be fairly easy to adjust, once the methods were known (but, isn't that always the case?!). I suspect that in an instance where the fading wasn't so definitively marked, corrections would have been more problematic. All in all, I was happy with the end result for this corrected photo.

PHOTO-SPECIFIC INFORMATION

NAME OF PHOTO: Construction Site/Dust/Hair

PROBLEM(S) TO BE CORRECTED: Remove hair and dust.

CORRECTIVE MEASURES TRIED WITH VARYING DEGREES OF SUCCESS:

ACTION: I began with the rubber stamp tool to clone window panes in the large archway with the hair running through it. Because the area was so small, however, it was difficult to clone only the areas I needed.

RESULT: I was constantly getting window frames where I didn't want them, or dark areas covering the frames. Though this could be a viable technique, I decided to try something else.

ACTION: Next I experimented with the magic wand tool. By double-clicking on the tool icon, I specified selecting like colors within 30 pixels. I then clicked on the hair.

RESULT: This didn't quite encompass all the hair I wanted, so I added to the selection by choosing "grow" and "similar" from the Select menu. I then filled this selection with the window black. The hair was still visible, just a different color.

ACTION: I undid my last action, and tried the magic wand tool again. This time I specified selecting like colors within 40 pixels. I then clicked on the hair.

RESULT: Practically all the white spots in the photo detail were selected, as a result. Of course, they didn't all need to be filled with the dark window color, so I selected undo.

ACTION: I then moved to the air brush tool in combination with a variety of brush sizes.

RESULT: I managed to fill in the hair in the archway successfully using the dark window color.

ACTION: Next, using the lasso tool, I selected, copied and pasted window frames to fill in missing areas of frame.

RESULT: The beauty of copying and pasting was not lost on me. It worked very well. Any sharp areas left after pasting could be easily corrected by blurring, or with a little touch up using fine brushes.

ACTION: I continued using copying, pasting, blurring, sharpening, and painting touch ups by toggling back and forth between eyedropper and small paint brushes.

RESULT: Small dust spots were very easy to get rid of. Missing stones, steps and ledges were easily copied and pasted. The most difficult areas to correct were in the intricate iron work of the crane, and the stair forms at the base of the crane. This was because of so much constant change in tone, and very fine lines to reproduce.

MOST SUCCESSFUL (OR ULTIMATELY APPLIED) CORRECTIVE MEASURE(S): Fine brushwork for details; copying and pasting; selective blurring.

GENERAL OBSERVATIONS: I was working on this photo simultaneously with the "Kids/Abrasion" and "Kids/Variegated Fading" photos. Compared to the two "Kids," I found the "Construction Site/Dust/ Hair," more enjoyable at first because I got a very dramatic positive result without the smudging I'd achieved on the "Kids/Abrasion" photo. It was very painstaking though. Also, because this was one of the first photos I worked on, I think I spent too much time on it. I corrected magnified dust spots that you never would have seen once the photo was reduced back down to its original size. In some ways, even though something like this is so detailed and intricate, it is easier to correct than something with a more gradual, smooth gradation of tones. You are less likely to see bad transitions.

PHOTO-SPECIFIC INFORMATION

NAME OF PHOTO: Construction Site/Variegated Fading

PROBLEM(S) TO BE CORRECTED: Adjust tones in over- and under-exposed areas of print so that bands of differing color are no longer discernible.

NOTE: In the process of working on "Kids/Variegated Fading," Natalie Munn suggested a technique that she had seen someone use. She thought I should try breaking the photo into three parts along its changes in exposure, that is, save it as three different documents (light, medium and dark parts). In theory, after adjusting the exposures of the dark and light bands by getting the Histogram readings to exactly match that of the medium band, one could paste the parts back together with absolutely no seam showing. As I had ended up using lassoed selections within one image for the "Kids" version, and since there was so much detail in the "Construction Site" photo, I thought it might be nice to try this technique and with luck avoid having to correct lines running through all the small items in the Construction photo.

CORRECTIVE MEASURES TRIED WITH VARYING DEGREES OF SUCCESS:

ACTION: Using the cropping tool, I broke the photo into three different documents.

RESULT: After first trying to paste one piece to the far right of another piece, but not being allowed by Photoshop to place an image flush right or left beyond an existing image, I discovered the Canvas Size command in the Image menu. This allowed me to add work space around an existing image. I specified the width and height I wanted in the dialog box, and indicated where I wanted the existing image to be placed.

ACTION: Having satisfied myself that I could get the pieces back together eventually, I proceeded to try to correct exposure by reading the Histogram information for the medium band. This was what I had set out to do originally for the "Kids" version - but now I understood that I had to try to get the same information to show in the Histograms of both the dark and light bands by using the same techniques of adjusting Curves, Levels, and Brightness/Contrast that I had for the "Kids/Variegated Fading" photo. However, since I had broken the "Construction Site" into three separate documents, I couldn't get them close enough together (because of intervening elevator bars, etc.) to judge by eye if the exposures truly matched. P> RESULT: I was constantly shifting from adjusting exposure, to going to Histogram to see if I was getting consistent information. I never managed to get the "Mean," "Std Dev," and "Med" levels of the Histogram to coincide for all three bands. It became quite frustrating.

ACTION: I decided to try the same technique I had used for "Kids" and hoped that the lines left between bands of corrected exposure would not be too difficult to eradicate.

RESULT: This seemed to work well. The lines left between bands of corrected exposure were not as noticeable as those on "Kids," either due to improved selection on my part, or more amenable tones in the photo.

ACTION: I cleaned up the slight lines that remained between the former areas of dark, medium and light. using a variety of cloning, blurring, sharpening and painting techniques already mastered in the "Kids" photo.

RESULT: The lines disappeared for the most part. The printout actually shows more of them remaining than was visible on the computer screen.

ACTION: Finally, using the lasso tool, I selected the sky area in the upper right-hand corner of the photo and toned down its brightness by using the Brightness/Contrast command.

RESULT: This brought the sky in this area more into agreement with the rest of the horizon. I also copied and pasted some of the crane back over areas that had accidentally been sprayed by overzealous puffs of cloned sky.

MOST SUCCESSFUL (OR ULTIMATELY APPLIED) CORRECTIVE MEASURE(S): Adjusted Curves, Levels, and Brightness/Contrast; rubber stamp to clone areas; fine brushwork for details; some selective blurring, sharpening, darkening and lightening.

GENERAL OBSERVATIONS: I was surprised to find that alleviating the remaining lines between corrected bands of exposure was no more difficult in this instance than in that of the smoother, less detailed "Kids." Because the difficulty in correcting these lines was not extremely high, I'm not sure that it would ever pay off to break a document into separate pieces that are then hard to judge side by side for color correction. I plan to discuss this more with Natalie.

PHOTO-SPECIFIC INFORMATION

NAME OF PHOTO: Missing Text/Emulsion

PROBLEM(S) TO BE CORRECTED: Repair missing areas of emulsion, and in so doing, make some moral determination about changing the content of the information presented

CORRECTIVE MEASURES TRIED WITH VARYING DEGREES OF SUCCESS:

ACTION: First I had to decide what I wanted to achieve. I thought about filling in the missing gaps of text, but lightening them, or somehow otherwise indicating that they were restorations to the original - ` la the patches of restored painted plaster in the Palace of Minos at Knossos on Crete. However, I decided to have some fun, and try to make my replacements as realistic as possible. By taking this path, I am well aware that the resulting photo is in no way a true recreation of the original. It merely indicates what an undamaged photo of the book might look like, not what it might say. I began by noticing that the illuminated word used throughout the text is the consistently same. This made it quite easy to copy parts of it, then paste and rotate to the appropriate angle in the two areas where it needed replacing.

RESULT: This seemed to be quite successful. From the Select menu I used the Feather command to soften copied edges, and the Defringe command to get rid of any remaining incorrectly colored pixels around the copied edges. I then used blurring and sharpening filters if necessary, in addition to various brush/eyedropper combinations to obscure the pasted areas.

ACTION: I next tackled small areas that needed repair. Where I could, I replaced missing text with the correct letter(s) if I could figure out what they were. I did this by copying, and pasting. I also color corrected with cloning, and with the dodging and burning tools as well as with some brush work. Gutters and decorative areas were cloned where possible. Creases in the far left margin were also cloned over with parchment texture.

RESULT: These repairs were generally successful. Using blurring filters helped blend everything in. Sometimes I had to rotate letters to get them to match the angle of the new word into which they were pasted.

ACTION: Finally I went for the large chunk of missing text on the right-hand side of the page. I began by selecting the big white area using the magic wand tool. Next I copied an undamaged area of text and tried pasting it into the selected area.

RESULT: The copied text didn't fill the white area completely. I undid the paste command.

ACTION: I then tried to save the selected "hole" as a mask into which I could paste text.

RESULT: I must admit that this process still seems confusing to me, and I didn't pursue it very far. I decided to go on with general copying and pasting instead.

ACTION: I proceeded to copy, paste and color correct as previously described. Finally, to make sure that the left-hand margin of the most damaged paragraph was even, I adjusted the spacing of words and letters falling along its newly created edge. I using the eyedropper tool to select the background color to match that of the parchment.

RESULT: When I moved words and letters to even out the spacing, the background color of parchment was revealed. This made it easier to cover with cloned areas of parchment and to smooth apparent edges by painting, lightening, darkening and cloning. A judicious blurring was added as the final touch to blend everything in.

MOST SUCCESSFUL (OR ULTIMATELY APPLIED) CORRECTIVE MEASURE(S): Copying, pasting and moving; rotating; rubber stamp tool to clone areas; fine brushwork for details and blending; blurring. Some dodging and burning. GENERAL OBSERVATIONS: I think I enjoyed the challenge of this problem the most of all that I worked on. It was interesting to guess what had been in the areas of lost emulsion and to try to match letters. Where possible, I tried to replace missing areas correctly. The fact that this problem is relatively easy to correct does give some pause, however. It would be quite possible to pass something off as authentic not only in appearance, but in content.


Impact Main Menu