Continuous Archiving of Group Digital Photograph Collections with a MPEG-7 Based Crowd Sourcing Annotation Methodology

Abstract

Following the increased convenience and availability of personal recording devices, the number of photographs and video clips one possesses can be multiplied at an accelerated speed. As a result of owning such a huge amount of digital data, one may encounter the difficulty of locating a specific photograph or video. Reformatting of the computer system or damaging on the hard disk also cause a potential risk of losing valuable personal data.Kuo, Aoki and Yasuda proposed an experiment personal archiving and retrieving image system called PARIS (Personal Archiving and Retrieving Image System) [2]. In PARIS, A MPEG-7 based multimedia description schema with extended spatial and temporal attributes is proposed to annotate personal multimedia data. While PARIS, the proposed experiment system, is designed and implemented specially towards the trend of continuous capture and storage for personal experience; it did not manage to utilize annotations that might come from recent emerging social networking enabled services.As a continuing project of PARIS, we initiated an experiment platform with thirty smart phone users. At the first stage, which last around six months, we encourage users to capture life events with smart phone cameras and to accumulate related photographs and video clips into our experiment database. In addition, we plan to extend this experiment with another 60 participants, which equipped with personal recording devices such as consumer digital cameras or digital video cameras in the near future.In this experiment database, we encourage users to upload their media files and provide related annotations. While we provide annotation options with previous proposed MPEG-7 based multimedia description schema, the new system generates annotation suggestions semi-automatically according to various on-line photograph-hosting services.Known as social-tagging, our proposed system allows users and visitors to create tags based on our previous proposed MPEG-7 structure and utilize pre-defined spatial and temporal ontology as well as resources provided via various on-line crowd generated resources. Users are allowed to create free personal tags and utilize our system generated suggestions in order to create relevant annotations which can lead to increased retrieval precision.