Image-Based Retrieval and identification of Ancient Coins

Image-based retrieval and identification of ancient coins at the Fitzwilliam Museum in Cambridge, England, was conducted by deviation from circular shape matching (DCSM) method. On the basis of the segmentation of the object, the border of the coin in the thresholded image were traced and a list of border pixel was obtained. Various descriptors such as scale-invariant feature transform (SIFT) and speeded-up robust features (SURF) emphasized different image properties such as intensity, edges, or texture of a coin. System workflow used the segmented region to obtain relevant features from the coin image and clicking on each filename listed in the center shows each image's dissimilarity score. The results show that SIFT clearly outperforms SURF independent of the size of training set and matching strategy. The quality and size of the training set is found to definitely influence the coin identification performance.