Adaptive histogram equalization: a parallel implementation

Abstract

Adaptive histogram equalization (AHE) is computationally expensive, and therefore time-consuming. A parallel method implemented with commercially available hardware is discussed. The implementation and testing conducted supports the hypothesis that multiple instruction/multiple data (MIMD) parallelism will improve the performance of the AHE algorithm. A method for handling border regions has been developed and implemented. The application of a look-up table served well to reduce the number of required computations. Whereas this probably did not affect the outcome of the testing, as both the sequential and parallel versions utilized this method, it did serve as an improvement to the overall implementation.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>