ABSTRACT
Where does patch come from? The earliest appearance of patch as a technical term in the open literature can be found in Reference [1], which focuses on textural features for image classification. In those old days, pixels are termed as resolution cells and small-area patches typically refer to a window or patch of 3×3 pixels. Later, Reference [2] introduced an image enhancement approach based on local statistics, such as mean and variance, calculated on a patch-by-patch basis in order to drive the filter to remove additive or multiplicative noise. Computing optical flow [3] or disparity field [4] from a pair of images was also based on the assumption that the gradient of image intensity field can be locally calculated within a patch, where patch is often interchanged with the term window or block. Fast advances of communication and computing technologies stimulated the research into image compression, especially the development of Joint Photographic Experts Group (JPEG) coding standard [5]. Under the context of JPEG compression, an 8× 8 discrete cosine transform (DCT) [6] became the favorite but also contributed to the notorious block artifacts; from this perspective, decomposing an image into non-overlapping patches or blocks represents a simple yet effective way of overcoming the curse of dimensionality [7].
