Thomas S. Huang

A fast two-dimensional median filtering algorithm

We present a fast algorithm for two-dimensional median filtering. It is based on storing and updating the gray level histogram of the picture elements in the window. The algorithm is much faster than conventional sorting methods. For a window size of m × n, the computer time required is 0(n).

The importance of phase in image processing filters

We demonstrate that phase accuracy is extremely important in image processing filters and express the hope that more work will be done on the development of filter design techniques which use phase as well as magnitude specifications.

Coding of Two-Tone Images

We review the concepts and techniques of efficient coding for the transmission or storage of two-tone images, such as business documents and weather maps. Digitization considerations are presented first. Then we discuss in detail the basic concepts (both heuristic and mathematical) of efficient coding. This is followed by descriptions of typical coding techniques. Finally, channel errors and buffer requirements are discussed briefly.

The effect of median filtering on edge location estimation

Abstract We study the effect of noise reduction preprocessing, specifically median filtering and averaging, on the accuracy of edge location estimation using least squares. The original edge is either a step or a linear ramp, corrupted by white Gaussian noise or binary symmetrical channel noise. The surprising conclusion is that in the case of white Gaussian noise, neither median filtering nor averaging improves the estimation accuracy. In the case of binary symmetrical channel noise, median filtering does improve the estimation accuracy for ramp edges, which are reasonable models for real-life edges.

Data compression for check processing machines

A technique of compressing image data derived from personal checks which possess several gray levels is described. Check images consist of both essential information such as printed and handwritten characters and nonessential background pattern or picture. Only the character plane is to be coded. Our proposed technique is divided into two phases: character plane extraction and character plane coding. In the first phase, a character plane which is composed of character pels on a uniform background is extracted from an original digital check image by using a combination of fundamental techniques of image segmentation. In the second phase, the extracted character plane is separated into a bit plane and a gray-level plane. The bit plane which preserves the position information of character pels on the character plane is conditional entropy coded. An adaptive two- or one-dimensional predictive coding scheme is applied to the gray-level plane which consists of only the character pels on the character plane. The check data are stored for further use as a combination of the codes derived from the bit plane encoder and the gray-level encoder in a check processing machine. A comparative study shows that the proposed coding scheme performs much better than conventional predictive coding schemes. For 8 gray-level image data, a compression factor of about 8:1 has been achieved.

Stochastic grammars and languages

This paper summarizes some recent results concerned with the extension of formal languages to their corresponding stochastic versions. Weighted grammars and languages are first defined, and stochastic grammars and languages are defined as a special case of weighted grammars and languages. Fuzzy grammars and languages, which have some properties similar to weighted grammars and languages, are also discussed. Stochastic automata are defined from the language recognition viewpoint. Languages accepted by stochastic finite-state and pushdown automata, with and without a cutpoint, are studied. Weighted and stochastic programmed and indexed grammars, and stochastic nested stack automata are defined. Finally, some decidability problems of stochastic (weighted, fuzzy) languages are discussed, and problems for further research are suggested.

Finding point correspondences and determining motion of a rigid object from two weak perspective views

The authors present a method which reduces an n-point problem to a set of four-point problems. The effort of reduction takes O(n) steps and it also takes O(n) steps to construct all possible mappings of an n-point set from the solution to a four-problem. Other results include (1) coplanarity condition of four points in two views, (2) recovering the tilt direction of the rotational axis using four points in two views, and (3) recovering the scaling factor.<<ETX>>

A subgraph isomorphism algorithm using resolution

Abstract An efficient algorithm for subgraph isomorphism is presented. It combines tree search with relaxation by using resolution. Bitwise parallelism, which is an important factor in speed, is achieved during the resolution process even though a sequential computer is used. The algorithm can be easily modified to apply to multi-relation labeled graphs, attributed graphs and some higher order structures such as arrangements.

A Syntactic-Semantic Approach to Image Understanding and Creation

A syntactic-semantic approach to information extraction from images is described. The methodology involves the injection of semantic considerations into context-free grammars. The semantic considerations include feature vectors, selection restrictions, feature transfer functions, semantic well-formedness, etc. With such injection, we can make a description scheme which carries the numerical, the structural, and the a prior real world knowledge about the pattern we want to extract. From the description we can construct an analytical mechanism, the creation machine, which wil find the desired pattern amid a chaos of noisy primitives.

An efficient algorithm for bilinear transformation of multivariable polynomials

An algorithm for bilinear-transforming multivariable polynomials is presented. It is similar to but simpler and computationally more efficient than existing algorithms.