A new image feature descriptor based on region edge statistical

Histogram is one of the most popular image features which are widely used in computer vision. However, lacking of space information in traditional gray histogram leads to an unsatisfactory performance in discriminating image features. To resolve this problem, this paper introduces the region edge complexity of histogram' subinterval and proposes a new method to describe the image features, namely, region edge histogram. Firstly, it defines a new concept of region edge complexity, and weighs each of the histogram bins using its corresponding region edge complexity. Then, it simplifies the expression of the weighted histogram and derives a compact form. Lastly, it presents an accelerated method to compute the region edge histogram based on integral histogram. The experimental results indicate that region edge histogram is superior to gray histogram and its ameliorations in feature discrimination and robust to the image rotation. Besides, it improves the efficiency at least two orders of magnitude than spatial histogram.