home
contents
copyright
donations
contact







PART III : POSTPROCESSING



Whereas Part I of this book offered advice on choosing photographic equipment and Part II provided guidance on the effective use of that equipment in the field, Part III deals with the processing of the resulting images on your computer.  We will consider not only the use of special effects to improve the visual appeal of your images, but also methods for organizing and maintaining the large corpus of image files that as an active digital photographer you’ll eventually accumulate.  While there are many books and tutorials available on the topic of image processing in Photoshop, our treatment in this part of the book will be quite specifically tailored to addressing the problems that most commonly arise when processing digital photos of birds.




Chapter 10

Fundamentals of Image Processing

Before diving in to a detailed discussion of the techniques for digital processing of bird photos, there are a number of important concepts about computer graphics that need to be understood.  Advanced readers will no doubt have a firm grasp on most of these concepts already, but for bird enthusiasts only recently introduced to computers, this chapter should fill in the relevant gaps about the basic underlying technology of digital image representation.  In order to articulate these concepts in as simple and concrete a manner as possible, we’ll assume the editing software to be used is the industry standard, namely Adobe Photoshop; for readers using alternative products, the underlying concepts will remain the same, though the details of the software interface may differ superficially.

10.1 Zooming, Cropping, and Resolution

Perhaps the first question needing to be addressed is why digital photos need to be processed at all.  In the days before digital photography, the film from your 35mm camera obviously needed to be processed (chemically) in order to produce a photographic print.  Now that cameras record their images directly to digital media, you might very reasonably expect that those images should be ready for immediate display on a computer (perhaps after being transmitted over the internet via a web page, or via email) or for sending to a digital printer. 
    Of course, one of the advantages of digital photography is that any blemishes or defects you perceive in the photo as captured by the camera can be corrected via software, and this is one good reason for learning about techniques for digital image processing.  But there’s a more fundamental reason—namely, the issue of target media.  In order for an image to be optimally rendered (i.e., displayed) on a medium such as a computer
s LCD screen or a sheet of photographic paper, the image file needs to be transformed into a format appropriate for that medium.  This transformation may involve mapping between color spaces (see section 14.1), mapping between computer platforms such as PC and Macintosh (via gamma correction—section 16.2), or simply adjusting the resolution of the image so as to be rendered at a desired size (i.e., on photographic paper of particular dimensions, or within a web page with a given formatting).  In this section we’ll deal with issues stemming from the latter consideration: resolution.



Fig. 10.1.1: The relationship between the image in memory and what is rendered on
the computer screen.  Top: a 10 megapixel image (left), when rendered on a 1 megapixel
screen (right) obviously results in a loss of some information, since each screen pixel
represents multiple image pixels.  Bottom: a magnified view of the raw image (left) shows
that the bird’s head retains considerable detail in the underlying photograph, while a
magnified view of the corresponding portion of the computer screen (right) shows
that those details can’t be rendered on the screen because the screen doesn’t have
enough pixels.  To see these image details on screen, you’d have to zoom in
and look at just a portion of the image.

    A useful mental exercise is to consider how a 10 or 15 megapixel image from a digital camera can be displayed on a roughly 1 megapixel screen—i.e., a typical 1440
×900-pixel computer monitor or laptop LCD (1440×900 = 1296000, or roughly 1 megapixel).   Obviously, since not all of the pixels in the image can be displayed on the screen at once (since the screen has fewer pixels than the image), you’ll have to either view just a part of the image, or somehow reduce the image resolution (number of pixels in the photograph) in order to fit the entire image on screen.  Today’s image-editing software makes this issue largely transparent by allowing you to zoom in and out of the image—i.e., to view the image at different resolutions, or zoom levels, so that different amounts of the image will fit on screen at one time.  This may at first seem a rather mundane issue, but there are some important subtleties that arise, particularly when digital manipulations (such as those described in the next chapter) are applied to the image and their aesthetic effect is appraised, by eye, at different zoom levels.
    Let’s first consider how the pixels comprising a digital image are mapped, by Photoshop or similar software, to pixels on your computer screen.  The cartoon below illustrates the basic concept.



Fig. 10.1.2: Interpolation of image pixels for rendering on a computer
screen.  If the raw image has more pixels than the screen, then displaying
the entire image on the screen all at once requires that each screen pixel
represent multiple image pixels.  The color of each screen pixel is an
interpolated value, or average, of the image pixels that it represents.
This is very important to keep in mind, as we’ll see later.

If the raw image has more pixels than your screen, then each pixel on your screen has to represent multiple pixels when the full image is rendered on the screen.  In the cartoon above, a block of 24 pixels, ranging in color from yellow to red, are highlighted at left.  Given the relative dimensions of the image and the screen in this hypothetical example, these 24 pixels in the image need to be mapped to a single pixel on the screen, in order for you to see the entire image all at once.  When the various yellow, orange, and red hues (precise color values) are averaged together numerically, they result in an orange hue, which is then used by the computer as the color of the single screen pixel representing these 24 pixels from the raw image.
    Now, if you were to apply a color adjustment in Photoshop, this orange pixel that you see on the screen would likely change to some other hue—perhaps to a somewhat lighter or darker hue of orange.  That’s what you’d see on your screen.  But because you’re not viewing the image at 100% zoom—that is, because each pixel on the screen represents more than one pixel in the image—you won’t see precisely how the image pixels are affected by the color adjustment you’ve just made.  All you’ll see is the average effect.  For the purpose of making color adjustments to an image, that’s typically not a problem.



Fig. 10.1.3: Zooming.  In order to fit the entire image on screen, you’ll generally have to zoom
out to some zoom level less than 100%.  For my particular computer screen and my particular
camera’s resolution, a 25% zoom level works best for me, but your hardware may be different.
Just keep in mind that sharpness and contrast (and some other image qualities) are best evaluated
at 100% zoom (possibly after re-scaling and cropping the image).  For detailed editing of
individual pixels, as with a digital paintbrush, a zoom level higher than 100% is often required.

    This issue becomes especially critical when applying image filters that affect the apparent sharpness of the image.  Because birds have so much fine detail in their feathers, sharpness is often one of the most critical issues when assessing overall image quality of a bird photo; this applies both to how other viewers assess your artwork and to how you decide which of your photos to publish, or which to keep and which to delete.  When postprocessing your images, you should always assess sharpness at 100% zoom (when each screen pixel represents one image pixel).  If you’re viewing an image at less than 100% zoom, you just need to keep firmly in mind that the apparent sharpness and contrast in the image, as you’re currently viewing it, may change dramatically when you resize the image to its final dimensions.  When you’re ready to begin tweaking the sharpness of the image, make sure you’re viewing the image at 100% zoom.  This is especially applicable for images that you intend to publish on the internet.  For images that you intend to print to photographic paper or canvas, I instead recommend an empirical approach involving iterative rendering to the target media until optimal settings are found—see section 14.1.


Fig. 10.1.4: Sharpening by the same amount at different zoom levels has a
profoundly different effect.  Left: sharpening at a radius of 0.15 has very little
visible effect when viewing the image at 25% zoom.  Right: after resizing
to 25% resolution and viewing at 100% zoom, sharpening at 0.15 radius
has an easily discernible effect.

    Largely separate from the issue of zoom level is the issue of cropping.  These two concepts are often confused, since zooming an image on screen can result in apparent cropping of the visible portion of the image in the viewing window.  The difference between zooming and cropping is simple: whereas zooming temporarily changes the apparent magnification and framing of an image on your screen, cropping permanently discards the margins of the image outside the crop region.  Zooming is something you’ll typically do liberally during postprocessing when switching between a full-image view to a pixel-for-pixel view of the image (and back again).  Cropping, on the other hand, is typically done just once to an image, to effect an artistic framing of the subject.  Cropping results in the actual deletion of pixels from the image, while zooming merely changes how the image’s pixels are displayed on your screen (and how many image pixels per screen pixel).
    Cropping is an important tool for artistic expression, since the manner in which an image is cropped—i.e., the precise dimensions of the cropped rectangle, and how those relate spatially to the original image boundaries—can profoundly affect the viewer’s perception of the subject and how that subject relates to its surroundings.  In section 8.1 we discussed composition principles for bird photos, which involved choosing the relative placement of the subject within the scene.  Implicit in that discussion was that the relative positioning of the subject relative to the scene was to be effected by movement of the lens prior to taking the photo—i.e.,
framing the shot.  In the field, composition effects are seen through the viewfinder of the camera, but during postprocessing you can often effectively modify the composition (i.e., the framing of the shot) via creative cropping of the image.
   


Fig. 10.1.5: When the subject is relatively small in the frame, cropping is often useful.
Precisely how you crop your images, both in terms of the re-scaling ratio and the
composition resulting from how you place the subject within the crop rectangle,
can profoundly affect the aesthetics of the final image.  I always explore a range
of re-sizing ratios and crop compositions in order to find the combination that
feels best to me.  Finding the best crop is sometimes the most time-consuming
part of my post-processing workflow.

    Since cropping permanently discards pixels from the outer margin of the image, it’s highly advisable to keep an original, unmodified version of the image in your archives in case you ever want to re-crop the image differently.  A prime example is if you first crop the image for posting on your web site, and then later decide to make a photographic print of the image.  When making prints, there are a number of circumstances in which you’ll want to include a larger margin around the subject than you would when posting snapshots on the internet.  One such example is when making canvas wraps (section 14.2); in the latter case you may need to include an additional 1 to 3 inches of margin for the sides of the wrap.
    Intertwined with the twin issues of zooming and cropping is the issue of resolution.  A number of definitions of resolution are in common use, and unfortunately, several different definitions are useful.  The main dichotomy, for our purposes, is between optical resolution and image resolution.  Optical resolution deals with the smallest details in the actual scene that can be resolved—i.e., visually separated, or discriminated, by the viewer—through an optical system.  For example, if you can confidently say that you can see the individual barbules in a bird’s feather, then you’re attesting to a higher degree of resolving power, or resolution, than if you can merely distinguish individual feathers from each other.  Optical resolution is useful to consider when choosing between different lenses.  When postprocessing images that you’ve already taken with your lens, this definition is less useful. 
    For existing images, resolution typically refers to the image dimensions, in units of horizontal and vertical pixel counts.  For the leftmost eagle image in the above figure, the resolution of the source image was 972
×648 pixels (width × height), whereas for the image on the right the resolution of the cropped source image was 566×377.  In this figure these two source images were rendered at the same print size of (approximately) 340×230 pixels.  This reduction in resolution was achieved in Photoshop via interpolation (i.e., averaging of adjacent pixel hues), with more aggressive interpolation required for the larger source image in order to reduce it to the same target size as the smaller image.  Because rescaling image pixels in this way typically reduces apparent sharpness, artificial sharpening in software is often performed after rescaling (see section 12.1).
    The practical importance of resolution differs somewhat for digital and print media.  For posting images onto web pages (i.e., on the internet), a reasonable range of sizes for rendering on today’s computer screens is from 400
×400 to 900×600.  The largest photos I post on my own web site are 1400×900, but these are so large that many viewers can’t see the whole image on their screen at once (keep in mind that the web browser program takes up some of the screen space with its menu, scroll bars, and navigation panes).  Different users on the internet have computers with different screen sizes, though most should be able to view a 600×900 image without difficultly (possibly after enlarging, or maximizing, their web browser window).  As LCD screen technology continues to advance, users’ screens will likely feature increasingly large numbers of pixels per inch, on average, so these numbers may change over time.



Fig. 10.1.6: Examining apparent detail at a given print size.  For this 30×20-inch canvas print,
I was glad to see that fine details were still apparent, even though the image was only 3600
×2400
pixels.  The ideal image resolution for a print task can depend both on the print medium and on
the individual characteristics of the particular photo.  A very sharp photo of lower resolution
may print better than a less-sharp photo of higher resolution.  Unfortunately, the only sure
way to know if your image is of high enough resolution is to print it on the target medium and
assess the quality of the final product by eye.

    For print media—i.e., actual photographic paper or canvas prints—much higher resolutions are usually desired in practice.  For example, my most popular canvas wrap measures approximately 30×20 inches, with a 1-inch border, and was made from a 3600×2400-pixel image (a mere 8.6 megapixels), though most of the 20×16 inch (plus margins) canvas wraps I’ve made were from 4800×3600 pixel images (~17 megaixels).  For 8×10 prints on standard photographic paper, I like to have an image resolution of least 2200×1800 pixels (~4 megapixels), though smaller images can sometimes still result in moderately decent prints.  The minimum image resolution for a given print size depends very much on the image itself, and can also depend on both the specific paper type and the printer used to make the print.  The only way to know for sure whether your image is of high enough resolution to render at a given print size is to actually make the print and assess the quality by eye.  More information on making prints is given in section 14.1.
    Adjusting image size in Photoshop is simple.  The Image Size dialog box lets you resize based on a desired pixel count (e.g., 900
×600), a desired scaling ratio (e.g., 50% of current size), or a desired print size (e.g., 8×10 inches).  For photos to be posted on a web page, simply bring up the Image Size dialog box, change pixels to percent in the drop-down menu, type the desired percent in the Width field (the Height field will be updated automatically to retain a constant aspect ratiothat is, ratio of width to height), and press the OK button.  I always use the Resample Image option with Bicubic as the resampling algorithm (the Bicubic Sharper doesn’t produce very good results, in my opinion—instead, I manually sharpen the image once before resizing, then resize it, and then sharpen again, using the Unsharp Mask, as described in section 12.1).



Fig. 10.1.7: Re-scaling an image in Photoshop.  The Image Size dialog box
allows you to scale an image by pixel counts, by document size (inches,
centimeters), or by scaling ratio (percent of current size).  For web
distribution of images, I typically rescale to either 33% or 50%.

    Now let’s briefly consider image navigation in Photoshop.  The figure below shows the various panes and palettes as they appear in the default layout of Photoshop.  Along the bottom you can see the current zoom level, which for this figure was set to 25% so that the entire image would fit on my computer screen.  At left is the main tools palette, from which you can select an individual tool to use in processing your image.  There are tools for cropping the image, adding text (such as a signature), painting (as with a digital paint brush), and selecting regions of the image (such as a bird’s head) that you’d like to process differently from the rest of the image; we’ll discuss all of these tools in detail later.  At the top of the screen, just below the main menu, is another pane that allows you to set the parameters of the current tool that you’re using—for example, the size of the brush that you’re painting with.  At right are the navigation and layers panes.  The layers pane shows you the layers making up the image.  In the figure below there is only one layer, but often you’ll want to isolate the bird by placing it on its own layer; this makes it very easy to apply effects to just the bird, or to just the background, thereby helping to make the bird stand out visually from its surroundings.


Fig. 10.1.8: Tools, menus, layers, and navigation panes in Photoshop.
The tools palette contains many useful tools that determine what happens
when you click or drag with your mouse cursor on the image.  The layers
panel shows image layers that you’ve created via the selection tools.  The
navigation pane is useful for panning around to different parts of the image
when viewing at large zoom levels.

The navigator is especially convenient when zooming in on specific regions of the image.  In the figure below, I’ve zoomed in to 100%, so that each pixel on my screen represents exactly one pixel in the image.  Because the image has a higher resolution than my screen, only part of the image can be viewed at one time.  In the navigator pane, you can see a red box around the bird’s head; this indicates the portion of the image that is currently visible in the main viewing area on my screen.  Using the computer’s mouse, I can easily pan around to different parts of the image by simply dragging that red box to the part of the image I want to see.  It’s an extremely efficient way of navigating around the image.



Fig. 10.1.9: Navigation in photoshop.  Panning to different parts of the image can
be achieved in multiple ways.  The navigator (upper right) shows you the full image
and allows you to pan by dragging the red box over the image.  The scroll bars allow
vertical or horizontal panning.  The hand tool allows you to drag the image around
within the main image window.  Finally, keyboard commands such as
Page Up
and Page Down can be used to navigate to different parts of an image.

An alternative to using the navigator is to drag the scroll bars (the thin blue segments shown along the bottom and rightmost margin of the main image window).  Yet another way to pan around in image space is to select the hand tool (from the tools palette at left) and to simply grab the image by pressing and holding the mouse cursor, and drag the image around within the main image window.  I find all three methods (the navigator pane, the scroll bars, and the hand tool) to be useful, at different times; once you get used to using all three methods you’ll find that you can navigate around the image very efficiently.
    Finally, let’s briefly survey the most useful tools available in the tools bar.  The figure below shows the tools that I use on a regular basis; those that I rarely or never use are grayed out.  You’ll notice that four of the tools are dedicated to selecting regions; as we’ll see in section 10.6, Photoshop provides a rich set of methods for selecting regions of the image that require special processing, and this will come in very handy for making the bird stand out from its background.  After the selection tools, the next most powerful and useful tool is the clone tool, which can be used either to remove unwanted elements from the scene (such as unsightly branches) or to repair blemishes, such as missing feathers (for birds in molt).  We’ll learn how to use this amazing tool in section 11.5.


Fig. 10.1.10: The toolbar in Photoshop.  The most useful tools for processing bird
photos have been outlined in red.  Most prominent are the selection tools,
which allow you to separate the bird from its background and then apply
different processing to each.  The clone tool is another extremely powerful
utility that you should work hard to master.  All of the tools have keyboard
shortcuts, which can drastically improve your efficiency.

The tool with the band-aid icon (the healing brush) is mostly useful for removing tiny spots due to dust on the camera’s imaging sensor.  Next to this is the brush tool, which paints pixels a solid color; this is mostly useful for fixing flash artifacts in the eye, and for adding a catchlight (see section 11.7).  The gradient tool (with the white-fading-to-black icon) is mostly useful for blending layers in a gradual manner, and will be discussed in chapter 13.  The part of the tools palette dedicated to changing brush colors is worth understanding, since it is useful both for fixing eye-shine and for adjusting selection boundaries in quick mask mode (section 10.6).  The white square represents the foreground color (white by default) and the black square corresponds to the background color (black by default).  Clicking on either of these will bring up a dialog box that allows you to select the new foreground or background color, while clicking on the tiny, curved arrow swaps the foreground and background colors.  The foreground color is the color that is used when you paint over parts of the image using the brush tool; the background color comes into play less often, though it can be used to quickly switch to an alternate painting color by clicking the swap colors arrow (or pressing the X key on your keyboard).
    The zooming (magnifying glass icon) and panning (hand icon) tools have already been discussed, though it’s worth noting that zooming can be achieved much more quickly by pressing the appropriate key combination—on Mac systems, the Cmd-+ and Cmd-_ combinations zoom in and out, respectively.  Some versions of Photoshop allow you to explicitly set keyboard shortcuts for all the tools, as well as for any of the special filters and other actions available via the menu, and I highly recommend taking full advantage of this feature, since it can drastically improve your efficiency when performing complex series of operations on an image.