Abstract

A binary object is a configuration of 1’s against a background of 0’s. Binary objects, such as arise in many industrial applications, may be particularly well restored by a Monte Carlo procedure. The problem is to find the correct configuration of the 1’s. First, how many 1’s are there? The number N of 1’s may be known to a good approximation by conservation of energy from the image data, especially when the image noise is Poisson. After estimating N, the algorithm places the N 1’s in an initial configuration consisting of the N highest-intensity values in the image. Following this, each 1’s position is randomly jiggled in x and y and the new position accepted if (1) it is previously unoccupied, (2) image inconsistency is reduced, and (3) a clump penalty term is satisfied (optional). The last-named condition takes advantage of a priori information that the object consists of isolated shapes, each having at least Nclump 1’s in each. Steps called merge and creep further augment the clump penalty. Applications to simulated imagery are shown.

© 1986 Optical Society of America

Origins of linear and nonlinear recursive restoration algorithms

Edward S. Meinel
J. Opt. Soc. Am. A 3(6) 787-799 (1986)

Restoration of images distorted by systems of random time-varying impulse response

Rabab K. Ward and Bahaa E. A. Saleh
J. Opt. Soc. Am. A 3(6) 800-807 (1986)

Restoration of images of partially obscured objects

Peter F. Jones and George J. M. Aitken
J. Opt. Soc. Am. A 14(5) 1015-1023 (1997)

Improved restoration of space object imagery

J. J. Green and B. R. Hunt
J. Opt. Soc. Am. A 16(12) 2859-2865 (1999)

Three-dimensional object recognition from multiple views

R. Wu and H. Stark
J. Opt. Soc. Am. A 3(9) 1543-1557 (1986)