Introduction
Images are a key element in web sites. Text-only web pages, like this one, are have only color, spacing, and typography to engage their visitors. (Actually, this one has the added element of “required reading for the course” to attract viewers to it.) But in general, images play critical roles in web pages: they provide design elements that help lift pages out of the drab and boring category, and they can also add content to the site in their own right. (If an article includes a photograph of the author, is that content or a design element? Sometimes the distinction isn’t clear.)
Sometimes you can get image files for your web site that are free and already suitable for what you want to use them for. For example, you can try Google Images or a photo sharing site, like Flickr. To avoid copyright problems, most sites with large audiences get their images either by hiring professional photographers or designers, or from “stock” sites, such as iStockPhoto. Stock images are not very expensive, and buying from a stock site guarantees no one will sue you for the use of the photo and/or publishing pictures of a particular person without their consent. The downside to stock photography is that somebody else could buy the same picture for use on their site. There have been some famous competing companies that used the same stock photograph in their print ads. The solution there is to pay more to the stock site, and they will give you exclusive use of a photo. (Or use your own pictures or hire a professional photographer/illustrator.)
But however you obtain the image files for your site, there is virtually no likelihood that you will be able to use the images “as-is.” They will have to be edited in various ways to make them “right for the web.” There are many excellent programs available for the editing task, including some very good ones that are free. The following list is by no means exhaustive; designer-info.com is a site that reviews most of these programs, and many others.
| Program | Notes |
|---|---|
| Photoshop | An expensive Adobe product that has many more features than what you need to prepare images for the web. But widely considered “the standard” program for editing image files. The topic of this web page, which introduces only a small subset of the program’s features. |
| Photoshop Elements | Has fewer features than the full Photoshop program, oriented towards amateur photographers. But it costs much less than the full version of Photoshop and may suffice for many web projects. |
| Fireworks | Another Adobe product, specifically designed for working with web images. Arguably better-suited for web designers than Photoshop, but not as well known. Originally developed by a company named Macromedia, which Adobe bought in 2005. (Macromedia also developed Dreamweaver and Flash.) |
| GIMP | The GNU Image Manipulation Program, an excellent and free alternative to Photoshop. |
| ImageMagick | A suite of command line tools suitable for preparing existing images for the web or other purposes. Because it works from the command line, it is great for automating image processing tasks. The “identify” command alone makes it worth installing the whole suite. |
| Inkscape | This is another excellent and free program, especially well-suited for web design work. Works in vector mode (see below) with SVG as its internal file format, but can export images as PNG files. |
| Illustrator | Yet another part of Adobe’s expensive “Creative Suite” of programs. This one is oriented towards vector graphics. |
Image File Formats
There are two ways to store images: vectors and bit maps. A vector image defines shapes by storing the
numerical coordinates of lines that make up the outline of the shape. Color fills and gradients are also
stored as numerical data as well. For example, this image: 
was created as an SVG file, which uses XML to represent lines, arcs, colors, and gradients.
Bit maps represent images as arrays of pixel values. For each pixel in the image, these formats store the red, green, and blue color intensities that will be mixed to produce that particular spot of light on the screen. Some bitmap formats also add a fourth “channel” of information to each pixel to specify its transparency. The image above is actually comes from a bit mapped image image file in the PNG file format. Browsers vary in their support of SVG image files: Internet Explorer 8 does not know what to do with them; Firefox 3 can display them as separate web pages; Opera, Chrome, and Safari can all display them using img tags.)
The advantage of vector drawings is that they keep sharp edges when they are resized. The disadvantages are that they cannot be used for (realistic) photographs and that they have to be converted to bit mapped images before they can be used in a web page. A good way to get a sense of the difference between vector and bit mapped graphics is to look at the web site for a program that will convert bit mapped images into vector images, Vector Magic.
The following table shows a few of the file formats that are used for images. At this time, only the first four can be used reliably in web pages.
| Acronym | Name | Type | Compression | Common Uses |
|---|---|---|---|---|
| JPEG | Joint Photographic Experts Group | Bit mapped; no transparency | Lossy | Photographs |
| PNG-8 | Portable Network Graphics | Bit mapped; up 8 bits of color information per pixel, used to index into a table of colors stored in the image; supports transparency | Lossless | Line drawings or photographs. |
| PNG-24 | Portable Network Graphics | Bit mapped; 24 bits of color information per pixel plus 8 bits of transparency information (“RGBA”) | Lossless | Line drawings or photographs. Also used by Fireworks as its “native” file format by adding metadata. |
| GIF | Graphics Interchange Format | Bit mapped; indexed colors; all–or–none transparency | Lossless | Line drawings |
| SVG | Scalable Vector Graphics | Vector | Lossless (if compressed) | Line drawings |
| TIFF | Tagged Image File Format | Bit mapped | Lossless | Photographs; can maintain information about image layers; JPEG is a special case of TIFF. |
| PSD | PhotoShop Document | Bit mapped | Lossless | Photoshop’s native file format. |
| AI | Adobe Illustrator | Vector | Lossless | Illustrator’s native file format. |
Image File Size
It is critical to edit image files so they will be as small as possible. The larger the image file, the longer it takes to download it from the server to the user’s browser. Without exception, a critical performance goal is to create web pages that download and display as quickly as possible, and image files are a major source of delays in loading web pages over the internet.
The size of a file is measured in bytes. One kilobyte (1 KB) is approximately one thousand bytes; one megabyte (1 MB) is approximately one million bytes; one gigabyte (1 GB) is approximately one billion bytes. So there are a thousand kilobytes in a megabyte and a thousand megabytes in a gigabyte.
There are three ways to reduce the size of an image file: reduce the number of pixels in the image, reduce the number of different colors that can appear at each pixel position, and remove redundant information that does not affect image quality.
Reduce the number of pixels
Image files have an intrinsic size.
For background images, the intrinsic size always determines how large the image will appear on the user’s screen. Repeating a background image horizontally and/or vertically can cause multiple copies of a background image to be shown, but each copy will have the image’s intrisic size, and there is nothing in CSS (or elsewhere) that you can use to change that fact.
Often, the intrinsic size of a background image is important for achieving the effect you want. For example, the background of this paragraph uses a 22 pixel–tall image to create a double horizontal line every 22 pixels; the intrinsic size determines the spacing between the lines. (The font size was set to 20px and the line-height was set to 1.1 (1.1×20 = 22) to get the text to sit evenly on the lines.)
But you can often reduce the intrinsic size of background images if they are going to be tiled by the browser. For example, the background image for the previous paragraph is only 1 pixel wide (that is, its intrinsic size is 1×22). The browser takes care of generating enough horizontal copies to create the line effect, as well as the number of vertical copies to fill the entire CSS box for the paragraph. The exception to this principle would be if you use a tiny image to tile over a very large area: depending on the speed of the user’s computer, the browser might need a lot of time to do the tiling.
For foreground images (using the img tag), the browser will display the image using its intrinsic size unless you specify a different size explicitly using the CSS width or height attributes.
-
If you drag an image into your web page using Dreamweaver, the resulting img
tag will have height and width
attributes added to it. Be sure to remove these attributes for two reasons:
- They violate the “best practices” policy that says to put all presentational controls, such as how big an image will appear, in the CSS stylesheet(s) for the document.
- Specifying both the height and width of an image is almost always a bad idea. The intrinsic size of an image also gives its intrinsic aspect ratio. If you change one dimension, you then have to change the other one in exactly the same proportion in order to prevent distorting the image. If you are going to have the browser change the size of an image from its intrinsic size, t is much better to set the size of just one dimension (either the width or the height) and let the browser calculate the other dimension using the image’s intrinsic aspect ratio.
-
If you make the image file’s intrinsic size bigger than it will appear on the web page, it will take longer to download than necessary.
Often, you will work with two copies of an image: the one you manipulate using your editing program, and the one you actually display on the web page. These two files will often use different formats, such as PSD for a Photoshop file that you use for editing and a PNG file that you export from Photoshop for use on your web page. The editng file might have a large intrinsic size because that is the way you got it or because it is often more convenient to edit an oversized image to achieve subtle effects and/or just to be able to see what you are doing. When your export the file for the web, be sure to specify a smaller intrinsic size to reduce the number of pixels in the image and thus the file size.
You can override the intrinsic size of an image using CSS, and the browser with enlarge or shrink the image as necessary. Shrinking an image will often make it appear sharper than it was before (the user can’t see the imperfections because they became smaller along with the rest of the image), but enlarging an image is never a good idea: the browser has to create more pixels than it has, and the new ones can only approximate what should have been there. In many cases, you get a “pixelated” image with jagged edges and/or fuzzy lines. So there is a trade-off here: if you make an image’s intrinsic size larger than needed, it can safely be reduced in size using CSS, but it will take longer to download than necessary.
Reduce the number of bits per pixel
There is some “information theory” to deal with here. The unit for measuring information is the bit, the amount of uncertainty that is reduced by answering one yes/no question. The two values a bit may have are 0 and 1. The number of bits needed to represent something depends on how many things might need to be represented. If a pixel can be only black or white, only one bit is needed (for example, 0 could mean black and 1 could mean white). But if a pixel can take on any of four colors (or shades of gray), it would take two bits to represent all four colors (for example, 00 could represent black, 01 could represent yellow, 10 could represent pink, and 11 could represent medium sea green). The good thing is that the number of possible colors doubles every time you add one bit: 3 bits can represent 8 colors, 4 bits can represent 16, 5 bits 32, etc. By the time you get to 24 bits (8 for the red intensity, 8 for the green intensity, and 8 more for blue), there are 224 different color possibilities—over 16 million! Computer displays and human eyes normally cannot discriminate among so many different colors. For example, can you see that the text in this sentence has a different color from the rest of the pargraph? (It has a color of #003311 compared to #003300 for the rest of the paragraph.)
Different image file formats use different methods to represent the colors of the pixels, but the general rule is that they use the same number of bits for every pixel in the image. If your image has just 32 colors in it, it needs just 5 bits per pixel. If you specify an image format that uses 15 bits per pixel, every pixel will occupy three times as many bits as necessary. The file probably will not be three times as large as it would be with 5 bits per pixel (there is other information in image files besides pixel data), but the larger the intrinsic size of the image, the bigger the waste.
Eliminate unneeded and redundant information
Non-image Information
Image files can contain non-image information that you can take out of the file when you save it for the web. Photoshop will take care of this for you when you use the “save for web and devices” menu item. But for other image editing programs, you may need to take care of this yourself:
- Color Profile. Color profiles tell a device how to translate a pixel’s numeric value into an actual color. The standard color profile for images displayed on the web is called sRGB, so there is no need to put the profile information inside the image file. Just be sure to convert any image files into sRGB colors when you save them if your image editor does not do so automatically.
- EXIF Data. Digital cameras put information inside the pictures they produce, including such information as the shutter speed and aperature, lens information, whether flash was used or not, etc. There is generally no need for including this information in pictures being downloaded to a web page, so it should be stripped out.
-
Other Metadata. Image files can also contain other non-image information, such as the name of the person
and/or program who created the image, how to contact the person, and possibly a copyright statement.
Except for photographs that belong to someone, it may make sense to remove some of this other information
in order to make the image file smaller. But never remove the copyright notice or contact information from
someone else’s photo before putting it on the web: doing so would make it look like you were trying
to defraud that person.
PNG files often make very good web images, but if you use Fireworks, be sure not to use the PNG file that program uses as its “native” file format; Fireworks adds tons of housekeeping metadata to its working files. Just use the Optimize option to export a copy of the file optimized for the web. Even if you just export the file as another PNG, Fireworks will strip out all its metadata from the copy, reducing the image file size very significantly.
Compression
File compression is another mechanism for removing redundant information from an image in order to make the file smaller. There are two kinds of compression: loss-y and loss-less. And example of lossless compression is a zip file: what you get out of the file by compressing it is exactly what was in it orginally, down to the last bit. An example of how lossless compression works to remove redundant information is run-length encoding. If a row of 500 pixels all have the same color, instead of repeating the same color number 500 times, the pixels would be stored as the one copy of the color number and a brief code indicating that it should be repeated 500 times. Loss-less encoding is a good thing, and many web servers will automatically zip files that it sends to a browser (provided the browser has indicated that it knows enough to un-zip them).
Lossy compression, however, is a different matter. JPEG uses lossy compression to reduce image file size on the presumption that the image will look good enough even if the picture quality is degraded by removing some of the possible colors and/or the number of separate pixel values in the image. And the experts (remember, the E in JPEG stands for “experts”) are right: many photographs can be subjected to a lot of lossy compression and still look good enough. Some factors affecting how much a picture can be compressed are how big its intrinsic size is (people cannot see as much detail in small pictures at normal viewing distances) and how the colors are distributed across the shapes in the picture. Both edges of objects, like flower petals or text, and expanses of sky are places to look for JPEG artifacts. Your editing program will let you specify the degree of compression you are going to use, while letting your preview the effect on the image. Normally, you enlarge the image on the screen so you can see the effects more clearly, but remember to do a reality check: view the image on the actual web page. You might be surprised how “bad” an enlarged image can be without showing any problems under normal viewing conditions.
Here are three copies of a picture saved using three different levels of compression:
daffodils-000
daffodils-030
daffodils-100
Viewed normally, you should be able to see that the first one has some blocks of color in the sky and some added colors in the petals. The Photoshop JPEG compression settings were 0 (most compressed; the number indicates “quality”), 30, and 100 for the three pictures. But look at the output from ImageMagic’s identify command:
Looking at the file sizes at the end of each line, you can see that the middle image is approximately 1/5 the size of the uncompressed version (10.8 KB / 51.1 KB). If you zoom in on the pictures (Firefox: Control +), you can see the JPEG artifacts in the middle image, but probably not at the normal viewing size (Firefox: Control 0).
Image Size Summary
-
Number of Pixels
Make the width and height match your needs. But it is okay to let the browser resize an image downward some, especially if the image is used in different sizes: it’s better to download one file and resize it in the browser than to download multiple copies of the same image differing only in size. Pixels per inch is irrelevant for web images. -
Bits per Pixel
The smaller the number of different colors, the smaller an image can be made. -
Eliminate Redundant Information
Take out unnecessary metadata. Use loss-less compression wherever possible. Use lossy (JPEG) compression with an eye on the actual image quality needed. -
Choose the Best File Format
When you use the menu item, Photoshop will give you a panel with a “four-up” view of the image, and various controls for manipulating the parameters listed above for JPEG, PNG-8, PNG-24, and GIF file formats.
There is also a re-size subpanel where you can change the intrinsic size of the image. Experiment with various alternatives in the three option views, using the original view as a referece point. When you get a combination that minimizes file size while retaining an acceptable appearance, you can proceed to save the file. Note that what you do here will have no effect on the original image file you are working with. You can make all the changes you want here, and still go back to editing the original image.
There is no simple rule for what file format to use: you have to experiment!