Real-time Image Sharing System For Smart Phones

In recent years, the number of online images has experienced an explosive growth with the rise and use of social image sharing sites. Due to the wide use of smartphones, more and more people are accustomed to sharing images. People tend to continuously take multiple shots for one scenic spot or one thing, and share all of them on the photo-sharing sites, which lead to repetition and approximation at these massive images in Internet. Meanwhile, the most current smart phones are limited by battery power. It must be a waste of a lot of power to share these repetitive or approximate. Therefore, it is important to share images in a real-time manner with the restriction of the energy of phones.This paper designed the real-time phone-sharing system for energy-limited smart phones, which supports fast image uploading while saving the battery power as possible as we can. The idea behind on the server side is to leverage space-efficient indexing structure and compact feature representation. With the compact feature representation, the images are transformed into feature vectors in the Hamming space. Locality Sensitive Hashing(LSH) is used in the indexing structure to support fast similar neighboring search by grouping similar images together. The conventional LSH unfortunately causes space inefficiency that is well addressed by the cuckoo hashing scheme. This paper takes advantage of a semi-random choice to improve the performance in the random selection of the cuckoo hashing scheme. An energy-aware module is implemented on the client side. The module can compress images according to the battery energy, then extract local keypoints and generate the feature vector. The energy-aware module uploads the battery energy to the server, and the final result is calculated and returned based on the energy on the server. Then, the client uploads the non-similar images.The prototype system is implemented, and several metrics are tested. The experimental results demonstrate the significant performance improvements.