1
E-STORE: An Energy-constrained Smartphone Storage for Near Real-time Disaster Image Sharing Pengfei Zuo, Yu Hua, Dan Feng, Zhenhua Nie, Min Fu, Yuanyuan Sun Wuhan National Laboratory for Optoelectronics, School of Computer Huazhong University of Science and Technology, Wuhan, China ■ Disaster environments — Images sharing for disaster relief ■ Challenges — Image redundancy — Energy constraint — Limited bandwidth ■ Existing schemes — Eliminate the redundant images in the forwarding path of network transmission — Overlook the energy constraint in smartphones ■ Energy-aware redundancy elimination in the source — Challenges: 1) High time and energy overheads for calculating image features; 2) The size of image feature is quite large, even larger than the image size — Solutions: 1) Energy-aware Dynamic Compression Scheme (Step 2); 2) A Conversion Algorithm (Step 3) ■ Fast query index for real-time response (Step 7) — Locality sensitive hashing: map the similar contents to the same bucket — Cuckoo hashing: deal with space inefficiency caused by LSH ■ Low battery —Energy-aware Threshold Setting Scheme (Step 8) ■ Large-size image compression before uploading (Step 11-1) — The high-quality images are not necessary for such disaster environments —Further reduce the bandwidth overhead ■ Evaluation configuration — Dataset: 50 images(60MB) — Emulate the network bandwidth in the disaster environments: 128Kbps — Redundancy ratio: from 0% to 100% ■ Preliminary results — 40% to 99.9% bandwidth saving — 33.9% to 93.8% time saving ■ Evaluate the performance of E-STORE using real-world datasets ■ Different network bandwidth and loads with a large number of smartphones ■ Measure and analyze the energy overhead of smartphones Background and Challenges The Proposed E-STORE System Preliminary Results Future Work 0 500 1000 1500 2000 2500 3000 3500 4000 4500 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Image upload delay (s) Redundancy rate Direct uploading E-STORE 0 10 20 30 40 50 60 70 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Network traffic (MB) Redundancy rate Direct uploading E-STORE The workflow of E-STORE 1. Select images and obtain the remaining energy of battery 2. Compress images based on the remaining energy 3.Extract the SIFT points and convert to image fingerprints 4.Upload the image fingerprints and the parameter of energy 5.Receive the data 6.Identify the similar images among the uploaded images 7.Query the image fingerprints in server index 8.Generate the threshold T depending on the parameter of the remaining energy 10.Do the similar images exist？ 11-1.Compress and upload the images 12.Receive the images 11-2.Do not upload Server Smart ph one 9.Respond with the query results No Yes
