A New Histogram Based Shape Descriptor in Image Retrieval

A shape based descriptor in image retrieval is proposed in this paper. It is focused on developing soft computing efficiency and novelty in image processing. The algorithm calculates a histogram based on the shape descriptor, representing texture feature of an image at high level (object oriented) effectively. By integrating the algorithms in key point detector with shape descriptor, the new method works fairly well compared with the stateof-the-art performance. The new detector detects relationships among key points, regardless of other pixels. That adds robustness to a large extent. Not only spatial relationships, but also variations in texture information are included in the descriptor as well. Structuring elements’ description (SED) and a set of overall texture descriptors in an image (short for TXT) are adopted for comparison. Experiments show that the new method performs the best with the 0.25 higher than the other two methods in robustness and accuracy. The new feature is flexible in multi-situations for different objects of interest.