Improved Bloom Filter for Efficient Image Retrieval on Mobile Device

A huge volume of images is uploaded to the mobile cloud environment every day with rapid growth of mobile devices. Different from retrieving image using a personal computer, infor-mation retrieval on mobile devices requires higher efficiency due to limitations in data transmission and memory cost. Therefore, an efficient scalable image retrieval model is proposed in this paper which consists of a two-layer bloom filter. The first layer of the proposed bloom filter generated from the asymmetric cyclical hashing (ACH) is used for primary image existence verification. In order to reduce the false positive rate commonly happening in bloom filters, the second layer of the bloom filter based on secure hashing is applied for verification in the second stage. The proposed model realizes approximated nearest neighbor retrieval with limited cost of storage and protects the stored data from illegal access simultaneously. Experimental results show that the proposed model yields significant better retrieval accuracy than other retrieval methods with more than 6 times faster in retrieval time and a smaller space requirement.