ABSTRACT
The goal of image compression is to store an image in a more compact form, i.e., a representation that requires fewer bits for encoding than the original image. This is possible for images because, in their “raw” form, they contain a high degree of redundant data. Most images are not haphazard collections of arbitrary intensity transitions. Every image we see contains some form of structure. As a result, there is some correlation between neighboring pixels. If one can find a reversible transformation that removes the redundancy by decorrelating the data, then an image can be stored more efficiently. The Karhunen-Loève Transform (KLT) is the linear transformation that accomplishes this.
