In this paper we address the comparison of two feature matching techniques which can be integrated in the HMAX framework. This comparison involves the originally proposed MAX technique and the histogram technique originating from Bag-of-Words literature. We have found that each of these techniques have their own field of operation. The histogram technique clearly outperforms the MAX technique with 5-15% for small dictionaries up to 500--1,000 features. A second investigation concentrates on comparing the often used hard vector quantization technique and a soft matching score technique for the histogram creation. It was found that the difference in performance is not significant and the scores are often within their standard deviations. Aiming at an embedded implementation such as in a surveillance system, computation power and memory (number of dictionary features) are intrinsically limited, so that the histogram technique is favored over the MAX technique.
|Title of host publication||Proceedings of the 30th Symposium on Information Theory in the Benelux, May 28-29, 2009, Eindhoven, The Netherlands|
|Editors||Tj. Tjalkens, F.M.J. Willems|
|Place of Publication||Eindhoven|
|Publisher||Technische Universiteit Eindhoven|
|Publication status||Published - 2009|