Data structures for quadtree approximation and compression

  • Authors:

  • Hanan Samet

  • Affiliations:

  • Univ. of Maryland, College Park

  • Venue:
  • Communications of the ACM
  • Year:
  • 1985

Quantified Score

Hi-index 48.22

Visualization

Abstract

A number of data structures for representing images by quadtrees without pointers are discussed. The image is treated as a collection of leaf nodes. Each leaf node is represented by use of a locational code corresponding to a sequence of directional codes that locate the leaf along a path from the root of the tree. Somewhat related is the concept of a forest which is a representation that consists of a collection of maximal blocks. It is reviewed and refined to enable the representation of a quadtree as a sequence of approximations. In essence, all BLACK and WHITE nodes are said to be of type GB and GW, respectively. GRAY nodes are of type GB if at least two of their sons are of type GB; otherwise, they are of type GW. Sequences of approximations using various combinations of locational codes of GB and GW nodes are proposed and shown to be superior to approximation methods based on truncation of nodes below a certain level in the tree. These approximations have two important properties. First, they are progressive in the sense that as more of the image is transmitted, the receiving device can construct a better approximation (contrast with facsimile methods which transmit the image one line at a time). Second, they are proved to lead to compression in the sense that they never require more than MIN(B, W) nodes where B and W correspond to the number of BLACK and WHITE nodes in the original quadtree. Algorithms are given for constructing the approximation sequences as well as decoding them to rebuild the original quadtree.