DOWE REALLY NEED GAUSSIAN FILTERS FOR FEATURE POINT DETECTION? Lee-kang Liu∗, Truong Nguyen

This paper studies the issue of which filters should be used for feature point detection. Classical feature point detection meth-ods, e.g., SIFT, are based on the scale-space theory in which Gaussian filters are proven to be optimal under the scale-space axiom. However, the recent method SURF demonstrates em-pirically that a box filter can also achieve good performance even though it violates the scale-space axiom. This leads to the question: Is Gaussian filters necessary for feature point de-tection? Based on the analysis using filter bank and detection theory, we show that theoretically it is possible for a box fil-ter to perform better than the Gaussian filter. Additionally, we show that a new filter, pyramid filter, performs better than both box and Gaussian filters in some situations. Index Terms — feature point detection, SIFT, SUR