Fast histogram equalization for medical image enhancement

To overcome the problem that the histogram equalization can fail for discrete images, a local-mean based strict pixel ordering method has been proposed recently, although it is unpractical for 3D medical image enhancement due to its complex computation. In this paper, a novel histogram mapping method is proposed. It uses a fast local feature generation technique to establish a combined histogram that represents voxels' local means as well as grey levels. Different sections of the combined histogram, separated by individual peaks, are independently mapped into the target histogram scale under the constraint that the final overall histogram should be as uniform as possible. By using this method, the speed of histogram equalization is dramatically improved, and the satisfactory enhancement results are also achieved.

I. Introduction

Histogram equalization is a widely used technique for image enhancement. When an image is equalized, its grey levels should spread till the margins of the entire scale. Also, the pixel number at each grey level should be made as close as possible. Thereby, the new image after equalization would possibly render better contrast effect for human vision, and details in some dark or bright regions could also show up. For continuous images, histogram equalization is relatively simple to complete, and owns exact solution. However, for discrete images, things become different. Usually, several minor grey levels would merge into only one grey level if the discrete images were equalized in the same way as the continuous images. More importantly, the details included in the minor grey levels will be lost after image equalization, although equalization itself is supposed to make details clearer after enhancement.