A Bayesian Matching Technique for Detecting Simple Objects in Heavily Noisy Environment
Authors :
Frantzeskakis, Emmanuel
Conference : SPIE 1991 International Symposium on Stochastic and Neural Methods in Signal Processing Image Processing Vol. 1569, pp. 341-353
Date: July 01 - July 01, 1991
The template matching problem for binary corrupted images with spatially white, binary, symmetric noise is studies. We compare matching based directly on the pixel-value image data as well as data coded by two simple schemes: a modification of the Hadamard basis and direct coarsening of resolution. Bayesian matching rules based on M-ary hypothesis are developed. The performance evaluation of these rules is provided,
The paper presents a study of the trade-off between the quantization level and the ability of detecting an object in an image. This trade-off depends on the (external) noise generated at the moment we receive the uncoded image.
The sum-of pixels and the histogram statistics are introduced in order to reduce the computational load inherent in the correlation statistic, with the resulting penalty of a higher probability of false alarm rate.
In the present work we demonstrate by examples that it is beneficial for recognition to combine an image coding technique with an algorithm extracting some “basic” information from the image. In other words coding (for compression) helps recognition. Numerical results illustrate this claim.
Download Full Paper