The histogram feature - a resource-efficient Weak Classifier

Abstract

Abstract-This paper presents a Weak Classifier that is extremely fast to compute, yet highly discriminant. This Weak Classifier may be used in, for example, a boosting framework and is the result of a novel way of organizing and evaluating Histograms of Oriented Gradients. The method requires only one access to main memory to evaluate each feature, in comparison with the more well-known Haar features which require somewhere between six and nine memory accesses to evaluate each feature. This low memory bandwidth makes the Weak Classifier especially ideal for use in small systems with little or no memory cache available.The presented Weak Classifier has been extensively tested in a boosted framework on data sets consisting of pedestrians and various road signs. The classifier yields detection results that are far superior than the results obtained from Haar features when tested on road signs and similar structures, whereas the detection results are comparable to those of Haar features when tested on pedestrians. In addition, the computational resources necessary for these results have been shown to be considerably smaller for the new weak classifier.

Boosting a heterogeneous pool of fast HOG features for pedestrian and sign detection

“…Pettersson et al [5] showed that a significant bottleneck for such systems is memory bandwidth. It was observed that memory access to the precomputed datatype for Haar features was faster than the memory bandwidth allowed.…”

“…Pettersson et al [5], introduced a novel precomputed image which they called the histogram image. The histogram image is used in a similar way to the precomputed images for other features, but has one major advantage in terms of speed.…”

“…The process of computing the histogram image is outlined in [5]. However, since most readers may not be familiar with its form, we provide an outline of the content of each histogram image pixel.…”

“…The use of precomputed datatypes, such as the integral image in [1], facilitates a much faster feature evaluation process than previous methods. In fact, evaluation requires so little processing time that these features are limited by bandwidth to memory, not CPU bandwidth [5]. In response to this problem, Pettersson et al [5] proposed a novel precomputed datatype, the histogram image, and a new feature, HistFeat, which is resource efficient in terms of memory and computation.…”

“…In fact, evaluation requires so little processing time that these features are limited by bandwidth to memory, not CPU bandwidth [5]. In response to this problem, Pettersson et al [5] proposed a novel precomputed datatype, the histogram image, and a new feature, HistFeat, which is resource efficient in terms of memory and computation. In this paper, we will present two feature types which both make use of the memory efficient histogram image, the LiteHOG and LiteHOG+ features.…”