This work first reviews an already-developed, existing deterministic parallel algorithm [2] to compute the complete histogram of an image in optimal number of steps on a hypercube architecture and utilizing memory space on the order of where x is the number of gray levels in the image, at each processing element. The paper then introduces our improvement to this algorithm�s memory requirements by introducing the concept of randomization into the algorithm.

image processing, computer vision, image histogram, hypercube architecture, randomized algorithms.