Improved local accumulate histogram-based Thangka Image Retrieval
Xiaojie Li, Weilan Wang, Wei Yang · 2010
This paper has accomplished an improved local accumulate histogram method of Thangka Image Retrieval. First, change the color space from RGB to HSV and divide similar area by the value of the hue reasonably, then it can get six similar intervals which the hue (H) is independent of the value (V) (Saturation (S) is assumed a constant in the beginning). Second, accumulate histogram is applied respectively for H, S and V in each local similar intervals, while S and V are obtained by mapping. Finally, the similarity of each local similar intervals are evaluated with Euclidean distance, before final retrieval result is obtained by setting weight for each result then re-measuring similarity. The experiment shows that this approach is effective not only in retrieving simple natural scenery images but also in retrieving Thangka images with rich color and complex texture.