We use the term image for still pictures and video for motion pictures or sequences of images. Image coding includes coding a wide range of still images such as bi-level or fax images, monochrome and color photographs, document images containing text, handwriting, graphics, etc.

Similarly to speech signals, images can be considered as outputs of sources with memory. In order to represent an image efficiently, it is important to take into account and remove any observable memory or redundancy. The typical forms of signal redundancy in most image and video signals are spatial redundancy and temporal redundancy, respectively.

Spatial redundancy takes a variety of different forms in an image. For example, it includes strongly correlated repeated patterns in the background of the image and correlated repeated base shapes, colors, and patterns across an image.

Temporal redundancy arises from repeated objects in consecutive images (frames) of a video sequence. Objects can remain or move in different directions. They can also fade in and out and even disappear from the image.

A variety of techniques used in the modern image-and video-coding standards to compensate spatial redundancy of images is based on transform coding considered in Chapter 5. The second basic principle of image coding is to exploit a human's ability to pay essentially no attention to various types of image distortion. By understanding the masking properties of the human visual system, it is possible to make distortions perceptually invisible.