LSB Elimination based Feature Extraction for Outsourced Image Retrieval in Encrypted Images

Abstract

This paper proposes an image retrieval approach in encrypted images that are outsourced to the cloud network using LSB (Least significant bit) elimination based feature extraction. To preserve the privacy of the image that is outsourced to the cloud, the images are encrypted. Retrieving the encrypted images that are outsourced to cloud is considered as a challenging task. The proposed method eliminates the LS B’ $s$ , and the histogram is estimated to obtain the necessary features. Initially, the image owner encrypts the original image by using a key and uploads the encrypted image to the cloud. The cloud server extracts the features from the encrypted original image and stores the features in the cloud. To perform image retrieval, the user has to encrypt the Query image using a key, and it must be uploaded to the cloud. The cloud server extracts the features from the Query image and matches with the features that are already stored in the cloud server. The encrypted images correspond to the top-k matched features are returned to the user, from which the user has to decrypt it using the key. The performance of the proposed method was measured in terms of Precision, time of feature extraction, time of index construction and time of image retrieval. Experimental results reveal that the proposed feature extraction outperforms than the traditional methods.