图像文件类型与格式指南

世界上有非常多的图像文件格式。下面列举了网页上常用的格式，其中 BMP 由于浏览器支持限制而不被推荐，通常应该避免在网页内容中使用。

每一种格式缩写都有一个详细说明，其功能以及详细的浏览器兼容信息，包括哪个版本引入支持以及以后可能会引入的特殊功能。

The following sections provide a brief overview of each of the image file types supported by web browsers.

In the tables below, the term bits per component refers to the number of bits used to represent each color component. For example, an RGB color depth of 8 indicates that each of the red, green, and blue components are represented by an 8-bit value. Bit depth, on the other hand, is the total number of bits used to represent each pixel in memory.

APNG is a file format first introduced by Mozilla which extends the PNG standard to add support for animated images. Conceptually similar to the animated GIF format which has been in use for decades, APNG is more capable in that it supports a variety of color depths, whereas animated GIF supports only 8-bit indexed color.

APNG is ideal for basic animations that do not need to synchronize to other activities or to a sound track, such as progress indicators, activity throbbers, and other animated sequences. For example, APNG is one of the formats supported when creating animated stickers for Apple's iMessage application (and the Messages application on iOS). They're also commonly used for the animated portions of web browsers' user interfaces.

The BMP (Bitmap image) file type is most prevalent on Windows computers, and is generally used only for special cases in web apps and content.

BMP theoretically supports a variety of internal data representations. The simplest, and most commonly used, form of BMP file is an uncompressed raster image, with each pixel occupying 3 bytes representing its red, green, and blue components, and each row padded with 0x00 bytes to a multiple of 4 bytes wide.

While other data representations are defined in the specification, they are not widely used and often completely unimplemented. These features include: support for different bit depths, indexed color, alpha channels, and different pixel orders (by default, BMP is written from bottom-left corner toward the right and top, rather than from the top-left corner toward the right and bottom).

Theoretically, several compression algorithms are supported, and the image data can also be stored in JPEG or PNG format within the BMP file.

In 1987, the CompuServe online service provider introduced the GIF (Graphics Interchange Format) image file format to provide a compressed graphics format that all members of their service would be able to use. GIF uses the Lempel-Ziv-Welch (LZW) algorithm to losslessly compress 8-bit indexed color graphics. GIF was one of the first two graphics formats supported by HTML, along with XBM.

Each pixel in a GIF is represented by a single 8-bit value serving as an index into a palette of 24-bit colors (8 bits each of red, green, and blue). The length of a color table is always a power of 2 (that is, each palette has 2, 4, 8, 16, 32, 64, or 256 entries). To simulate more than 255 or 256 colors, dithering is generally used. It is technically possible to tile multiple image blocks, each with its own color palette, to create truecolor images, but in practice this is rarely done.

Pixels are opaque, unless a specific color index is designated as transparent, in which case pixels colored that value are entirely transparent.

GIF supports simple animation, in which following an initial full-size frame, a series of images reflecting the parts of the image that change with each frame are provided.

GIF has been extremely popular for decades, due to its simplicity and compatibility. Its animation support caused a resurgence in its popularity in the social media era, when animated GIFs began to be widely used for short "videos", memes, and other simple animation sequences.

Another popular feature of GIF is support for interlacing, where rows of pixels are stored out of order so that partially-received files can be displayed in lower quality. This is particularly useful when network connections are slow.

GIF is a good choice for simple images and animations, although converting full color images to GIF can result in unsatisfactory dithering. Typically, modern content should use PNG for lossless and indexed still images, and should consider using APNG for lossless animation sequences.

The ICO (Microsoft Windows icon) file format was designed by Microsoft for desktop icons of Windows systems. However, early versions of Internet Explorer introduced the ability for a web site to provide a ICO file named favicon.ico in a web site's root directory to specify a favicon — an icon to be displayed in the Favorites menu, and other places where an iconic representation of the site would be useful.

An ICO file can contain multiple icons, and begins with a directory listing details about each. Following the directory comes the data for the icons. Each icon's data can be either a BMP image without the file header, or a complete PNG image (including the file header). If you use ICO files, you should use the BMP format, as support for PNG inside ICO files wasn't added until Windows Vista and may not be well supported.

The JPEG (typically pronounced "jay-peg") image format is currently the most widely used lossy compression format for still images. It's particulary useful for photographs; applying lossy compression to content requiring sharpness, like diagrams or charts, can produce unsatisfactory results.

JPEG is actually a data format for compressed photos, rather than a file type. The JFIF (JPEG File Interchange Format) specification describes the format of the files we think of as "JPEG" images.

The PNG (pronounced "ping") image format uses lossless or lossy compression to provide more efficient compression, and supports higher color depths than GIF, as well as full alpha transparency support.

PNG is widely supported, with all major browsers offering full support for its features. Internet Explorer, which introduced PNG support in versions 4–5, did not fully support it until IE 9, and had many infamous bugs for many of the intervening years, including in the once-omnipresent Internet Explorer 6. This slowed PNG adoption, but it is now commonly used, especially when precise reproduction of the source image is needed.

SVG is an XML-based vector graphics format that specifies the contents of an image as a set of drawing commands that create shapes, lines, apply colors, filters, and so forth. SVG files are ideal for diagrams, icons, and other images which can be accurately drawn at any size. As such, SVG is popular for user interface elements in modern Web design.

SVG files are text files containing source code that, when interpreted, draws the desired image. For instance, this example defines an drawing area with initial size 100 by 100 units, containing a line drawn diagonally through the box:

<svg viewBox="0 0 100 100" xmlns="http://www.w3.org/2000/svg">
  <line x1="0" y1="80" x2="100" y2="20" stroke="black" />
</svg>

SVG can be used in web content in two ways:

1. You can directly write the <svg> element within the HTML, containing SVG elements to draw the image.
2. You can display an SVG image anywhere you can use any of the other image types, including with the <img> and <picture> elements, the background-image CSS property, and so forth.

SVG is an ideal choice for images which can be represented using a series of drawing commands, especially if the size at which the image will be rendered is unknown or may vary, since SVG will smoothly scale to the desired size. It's not generally useful for strictly bitmap or photographic images, although it is possible to include bitmap images within an SVG.

TIFF is a raster graphics file format which was created to store scanned photos, although it can be any kind of image. It is a somewhat "heavy" format, in that TIFF files have a tendency to be larger than images in other formats. This is because of the metadata often included, as well as the fact that most TIFF images are either uncompressed or use compression algorithms that still leave fairly large files after compression.

TIFF supports a variety of compression methods, but the most commonly used are the CCITT Group 4 (and, for older fax systems, Group 3) compression systems used for by fax software, as well as LZW and lossy JPEG compression.

Every value in a TIFF file is specified using its tag (indicating what kind of information it is, such as the width of the image) and its type (indicating the format the data is stored in), followed by the length of the array of values to assign to that tag (all properties are stored in arrays, even for single values). This allows different data types to be used for the same properties. For example, the width of an image, ImageWidth, is stored using tag 0x0100, and is a one-entry array. By specifying type 3 (SHORT), the value of ImageWidth is stored as a 16-bit value:

Specifying type 4 (LONG) stores the width as a 32-bit value:

A single TIFF file can contain multiple images; this may be used to represent multi-page documents, for example (such as a multi-page scanned document, or a received fax). However, software reading TIFF files is only required to support the first image.

TIFF supports a variety of color spaces, not just RGB. These include CMYK, YCbCr, and others, making TIFF a good choice for storing images intended for print, film, or television media.

Long ago, some browsers supported TIFF images in web content; today, however, you need to use special libraries or browser add-ons to do so. As such, TIFF files are not useful within the context of web content, but it's common to provide downloadable TIFF files when distributing photos and other artwork intended for precision editing or printing.

WebP supports lossy compression via predictive coding based on the VP8 video codec, and lossless compression that uses substitutions for repeating data. Lossy WebP images average 25–35% smaller than JPEG images of visually similar compression levels. Lossless WebP images are typically 26% smaller than the same images in PNG format.

WebP also supports animation: in a lossy WebP file, the image data is represented by a VP8 bitstream, which may contain multiple frames. Lossless WebP holds the ANIM chunk, which describes the animation, and the ANMF chunk, which represents a frame of an animation sequence. Looping is supported.

WebP now has broad support in the latest versions of major web browsers, although it does not have deep historical support. Provide a fallback in either JPEG or PNG format, such as with the <picture> element.

XBM (X Bitmap) files were the first to be supported on the Web, but are no longer used and should be avoided, as their format has potential security concerns. Modern browsers have not supported XBM files in many years, but when dealing with older content, you may find some still around.

XBM uses a snippet of C code to represent the contents of the image as an array of bytes. Each image consists of 2 to 4 #define directives, providing the width and height of the bitmap (and optionally the hotspot, if the image is designed as a cursor), followed by an array of unsigned char, where each value contains 8 1-bit monochrome pixels.

The image must be a multiple of 8 pixels wide. For example, the following code represents an XBM image which is 8 pixels by 8 pixels, with those pixels in a black-and-white checkerboard pattern:

#define square8_width 8
#define square8_height 8
static unsigned char square8_bits[] = {
  0xAA, 0x55, 0xAA, 0x55, 0xAA, 0x55, 0xAA, 0x55
};

Picking the best image format for your needs is likely easier than video formats, as there are fewer options with broad support, and each tends to have a specific set of use-cases.

Photographs typically fare well with lossy compression (depending on the encoder's configuration). This makes JPEG and WebP good choices for photographs, with JPEG being more compatible but WebP perhaps offering better compression. To maximize quality and minimize download time, consider providing both using a fallback with WebP as the first choice and JPEG as the second. Otherwise, JPEG is the safe choice for compatibility.

For smaller images such as icons, use a lossless format to avoid loss of detail in a size-constrained image. While lossless WebP is ideal for this purpose, support is not widespread yet, so PNG is a better choice unless you offer a fallback. If your image contains fewer than 256 colors, GIF is an option, although PNG often compresses even smaller with its indexed compression option (PNG-8).

If the icon can be represented using vector graphics, consider SVG, since it scales across various resolutions and sizes, so it's perfect for responsive design. Although SVG support is good, it may be worth offering a PNG fallback for older browsers.

Unless you're willing to compromise on quality, you should use a lossless format for screenshots. This is particularly important if there's any text in your screenshot, as text easily becomes fuzzy and unclear under lossy compression.

PNG is probably your best bet, but lossless WebP is arguably going to be better compressed.

For any image that can be represented using vector graphics, SVG is the best choice. Otherwise, you should use a lossless format like PNG. If you do choose a lossy format, such as JPEG or lossy WebP, carefully weigh the compression level to avoid causing text or other shapes to become fuzzy or unclear.

While the standard HTML <img> element doesn't support compatibility fallbacks for images, the <picture> element does. <picture> is used as a wrapper for a number of <source> (en-US) elements, each specifying a version of the image in a different format or under different media conditions, as well as an <img> element which defines where to display the image and the fallback to the default or "most compatible" version.

For example, if you're displaying a diagram best displayed with SVG, but wish to offer a fallback to a PNG or GIF of the diagram, you would do something like this:

<picture>
  <source srcset="diagram.svg" type="image/svg+xml" />
  <source srcset="diagram.png" type="image/png" />
  <img
    src="diagram.gif"
    width="620"
    height="540"
    alt="Diagram showing the data channels" />
</picture>

You can specify as many <source>s as you wish, though typically 2 or 3 is all you need.

• Guide to media types and formats
• Web media technologies
• Guide to video codecs used on the web
• The HTML <img> and <picture> elements
• The CSS background-image property
• The Image() (en-US) constructor and the HTMLImageElement interface