H2: Fusion of HOG and Harris features for image segmentation evaluation
Macmillan Simfukwe, Bo Peng, Tianrui Li · 2017
Image segmentation is a vital task in image processing/computer vision. However, no universally accepted quality measure exists for evaluating the performance of various segmentation algorithms or even different parameterizations of the same algorithm. This paper proposes a new segmentation evaluation measure, based on the fusion of HOG and Harris features, thus we call it the H2. It exploits local shape, corner and edge information to evaluate the similarity between a given segmentation and its respective ground truth, and thus belongs to the category of supervised evaluation measures. The results obtained from our experiments show accuracy of up to 95% for the H2.