A study on low-cost representations for image feature extraction on mobile devices

Abstract

Due the limited battery life and wireless network bandwidth limitations, compact and fast (but also accurate) representations of image features are important for multimedia applications running on mobile devices. The main purpose of this work is to analyze the behavior of techniques for image feature extraction on mobile devices by considering the triple trade-off problem regarding effectiveness, efficiency, and compactness. We perform an extensive comparative study of state-ofthe- art binary descriptors with bag of visual words. We employ a dense sampling strategy to select points for low-level feature extraction and implement four bag of visual words representations which use hard or soft assignments and two most commonly used pooling strategies: average and maximum. These mid-level representations are analyzed with and without lossless and lossy compression techniques. Experimental evaluation point out ORB and BRIEF descriptors with soft assignment and maximum pooling as the best representation in terms of effectiveness, efficiency, and compactness.