Feature classification for representative photo selection

Abstract

This paper points out that different local feature points provide different impacts to near-duplicate detection and related applications. Aiming to automatic representative photo selection, we develop three feature classification methods, i.e., point-based, region-based, and pLSA-based classification, to differentiate local feature points described by SIFT descriptors. We investigate the performance of these classification methods, and discuss how they influence near-duplicate detection and extended applications. Experiments show that, with effective feature classification, more accurate representative selection results can be achieved.

Visual summarization for place-of-interest by social-contextual constrained geo-clustering

“…The mean shift algorithm is often utilized to group geotagged photos into clusters, from which representative images can be selected [1], [6][7][8].…”

“…In recent years, much work aim to mine POI worldwide based on these metadata. Some of the exiting works use a variety of methods to carry out representative pictures recommendation [1][2], [8], and the corresponding travel recommendations [6], [9][10].…”

“…And many research efforts have been dedicated to representative image recommendation. Many of them [2], [4], [6], [8] consider both visual and text features to solve their problems. For instance, Lyndon et al [2] used content-based methods to choose diverse and representative images for a given landmark.…”

“…Given a set of images with the specified tags, they generated canonical views by clustering of images visual properties, and extracted representative tags for each cluster. Chu et al [8] proposed a novel method to automatically select representative photos by feature classification. They mainly used the local features to distinguish artificial objects and natural objects, and chose the near-duplicate parts as the representative landmark.…”

Automatic summarization of travel photos using near-duplication detection and feature filtering

“…We improve this work by differentiating different types of features [2]. First, each photo is divided into 40 40 regions.…”