On Accuracy and Speed of Object Recognition Based on Local Arrangements of Feature Points

The geometric hashing is a well-known object recognition technique based on the arrangements of feature points. We have proposed “locally likely arrangement hashing (LLAH)” which outperforms the geometric hashing in both retrieval accuracy and processing time. In this report, by comparing both methods, we consider the major factors which bring the improvement. We also consider the relationship between a picture angle and the accuracy of the LLAH because the accuracy of the LLAH depends on the picture angle.