EditingGuidelines/Images

From J Wiki
Jump to: navigation, search

It is good to enhance text with graphical illustrations. Wiki makes it easy to add images by uploading them as attachments and linking in the text. With a few simple considerations you can make your screenshots and pictures look good on the page. Not all of them are necessary for every image you make, but it's good to be aware of different techniques useful for particular cases.

Common Sense

There are a few considerations behind making a good screenshot or illustraton.

  • should look good on screen; for example, screen is color, so take advantage of color, gradients, image effects if possible;
  • should print well; printer scales, so smooth antialiased contours will scale better than computer graphics with pixellation or jaggered lines;
  • should be the right size; huge images take up the whole screen, for good measure look at Print View and see that it fits page margins; too small pictures may not look good all by themselves, so consider a floating box in a paragraph; you can also combine a few images in a horizontal table
  • keep screen proportion; an ideal illustration is landscape taking abount 1/3 of screen space or less with proportions of the screen
  • balance brightness and contrast; printer may be black and white; all dark images do not look good and take much ink
  • level of detail should correspond to typeface and images around it; e.g. a chart with too large or too small font or too thick lines does not match

Taking a Good Picture

Here we will go over every step from preparing the subject of shooting to getting the picture ready to be published in some detial.

Preparing the View

  • if it is a generated image, see if you can balance the parameters so that the resulting image conveys enough information without overcrowding the view
  • resize the window before taking the screenshot; typically you want to resize it to the smallest size before distortion or cutting off happens; resize to desired proportion; resizing a plot, for example, watch the axes to obtain the optimum gridlines steps
  • resize the GUI screen to remove extra white space between controls; smaller size original screenshots do not require scaling, so they will look better
  • take screenshots of single windows by themselves (Alt+PrtSc in Windows), this will prevent picking up extraneous background; to take several windows at once, select a white background, e.g. by placing blank web browser behind, and take a full screen shot and crop
  • overlapped windows (such as here) could be taken by carefully positioning them, so that important bottom titles are exposed and most of unnecessary areas are covered; alternatively they can be combined in Enhancing stage with individual windows taken on white background, paying attention so that extraneous borders and other artifacts not sneak in

Enhacing Visualization

  • decide about scope of the image (small detail or fullscreen) based on what you want to show: styling of a button borders or a general view of how a GUI form looks like; in post-production it will be rescaled to fit the page
  • place arrows, ovals, underline, captions, etc. to guide the viewer; esp. useful in GUI how-to's
  • use empty spaces in irregular shapes for comments, chart, table, alternative view
  • combine several irregular shapes or a sequence of small images into one denser image, like here
  • combine two scales of the same picture with detail view or thumbnail view

Image Optimization

Finally you are happy with what your illustration contains, now it's time to put it into final form, ready for placing on the page.

The goal of optimization is to reduce the resolution and file size while preserving a sufficient amount of detail.

Cropping:: to cut out the required portion, you can either select and copy/paste to a new editor or crop in place (see note on Windows Paint below) Combining:: several different shapes or overlapped windows can be combined, such as in Grid Test Drive or Plot 3D; with some practice it is quick and easy to do in a simple editor like Paint using the clipboard, the empty space of the same image and Undo function. To overlap irregular shapes, in Paint deselect the Opaque mode Scaling:: to have reasonable size, some images need to be scaled. It is best to use ratios of powers of 2 such as 75%, 50% or 150%, etc, unless you use a high-end editor with non-linear scaling. Enhancing:: after rescaling, images may loose contrast or become too bright; sometimes sharpening is necessary; to reduce pixellation and jagged lines, consider smoothing or smooth &. expand Stylizing:: sometimes decorations like shadows, texturing or selecting a good type for title may create a better view, but these should be used with restraint Saving:: for line art images such as plots use PNG format; for smooth images, such as photos use JPEG. Compression should produce smallest size without significant distortion, which varies with different pictures and purpose: precision archive quality or casual overview screenshot Naming:: last but not least is a good naming scheme for the images, so that it is easy to find or distinguish them later; but with for 1-2 illustrations plot1.png, plot2.png is good enough.

Placing Images

For general Wiki guidelines, see HelpOnLinking, HelpOnActions/AttachFile and HelpOnSmileys.

Smaller image, on a line by itself, looks better if indented: insert a space before inline:.

Floating Boxes

Floating boxes allow to place a small image flush at the margin embeded within the paragraph text, such as a simple box or a box with a title.

||<tablestyle="float:right"> inline:jal36.png||
This is paragraph text the image is floating with ...

To remove border, add style="border:0" between < >

Horizontal Table

Several smaller images can be shown side-by-side using a table, such as an image strip or an image table

|| inline:img1.png || inline:img2.png || inline:img3.png ||

Image Editors

Different programs can be used for editing and enhancing images.

On Windows a good choice is the standard Paint program. It is getting subtly improved with every version. The one for Windows XP supports PNG and has a good scale-down (decent bilinear), which works good on multiples of 2 scales, e.g. 75%. It has solid basic editing too, including large scale for icons, such as small flags.

For higher quality image processing similar to Photoshop filters.

Information.png Free all-platform XnView with near-Photoshop adjustment, filter and export capabilities is highly recommended. It is very easy to use.

Paint has very solid feel of selecting, moving around fragments and cropping operations, opaque or transparent background modes. It is also very precise, e.g. to crop a view: select all, move to cut off top-left, deselect, drag the bottom-right corner to cut off the opposit part.

On Mac you can use the bundled AppleWorks with both drawing and painting programs. For image adjustment, iPhoto is a great tool.

Heavier weight, almost like Photoshop, but free is GIMP, good for vector paths, fonts, shadows and layer compositing.

Image Optimization Example

Here we will take a scan made at large resolution, and convert it to image that preserves all information and detail while fitting nicely on a page. We will use XnView.

Original: 1,000 Kb, 2464 x 2832
Optimized: 100 Kb, 887 x 1019

0. Obserivng the original, we see that it is too large for screen viewing and has residual shades on paper.
0 original.jpg

1. Reducing the size. We will use the screen font criterion: matching the font size on the image to Courier New 10pt on the screen. Selecting menu Image

Paint has very solid feel of selecting, moving around fragments and cropping operations, opaque or transparent background modes. It is also very precise, e.g. to crop a view: select all, move to cut off top-left, deselect, drag the bottom-right corner to cut off the opposit part.

On Mac you can use the bundled AppleWorks with both drawing and painting programs. For image adjustment, iPhoto is a great tool.

Heavier weight, almost like Photoshop, but free is GIMP, good for vector paths, fonts, shadows and layer compositing.

Image Optimization Example

Here we will take a scan made at large resolution, and convert it to image that preserves all information and detail while fitting nicely on a page. We will use XnView.

Original: 1,000 Kb, 2464 x 2832
Optimized: 100 Kb, 887 x 1019

0. Obserivng the original, we see that it is too large for screen viewing and has residual shades on paper.
0 original.jpg

1. Reducing the size. We will use the screen font criterion: matching the font size on the image to Courier New 10pt on the screen. Selecting menu Image | Resize we experiment a few times to find a good match (36%), by comparing width of the same sentence. We use Lanczos resampling.
1 scale.jpg

2. Among image adjustments we will use Contrast and Brightness. Choose from menu Image | Adjust | Brightness/Contrast... First increase contrast to remove dark residuals, then decrease brightness to match the density of glyph stems to that of a screen font in another window.
2 adjust.png

3. For export (optimizing resolution and size vs detail) we choose PNG to reduce quantization artifacts. Since there are only few colors, we choose indexed color, but to preserve font anti-aliasing we set the level to 64.
3 export.png

Adaptive Scaling

With the more recent version of MoinMoin, it is now possible to specify attributes of the underlying HTML elements. In particular the <IMG> element, has the attribute width="123" where 123 is the resulting size in pixels.

This allows maintaining an adaptive DPI resolution strategy: for the display images to fit on screen, yet preserving the image resolution for high density printing. Note: the file size matters and too large images are undesirable. A practical ratio of the screen/print scaling is from 33% to 66%.

Taking the above example from NYCJUG agenda scans using NYCJUG/2005-02-08 and applying the following attributes:

Template:Attachment:NYCJUGMeetingNotes20050208.jpg

Note: the magic number 980 is the width of a "standard" viewport for HTML pages chosen for iPhone. The original image width is the pixel size of the scanned page. The resulting ratio 980 % 2464 = 0.40 roughly corresponds to the ratio from the previous section.

we have the following result, where textual font size matched scanned font size.

Agenda.png