A Secured Privacy Preserving CBIR Using Color Co-Occurrence Matrix And Gabor Wavelet Transformation

Abstract

With rapid development of images across networks, Content based image retrieval (CBIR) is the predominating technique for search and retrieval of images over a large datasets. Since the images are outsourced to third party cloud servers, maintaining privacy of users and integrity, confidentiality of images is a challenging task. In our paper, we address the concern of privacy of images as it contains various sensitive information. Our proposed model first computes the features with respect to color using Color Co-occurrence Matrix and textures using the Gabor wavelet transformation. The generated features are mapped with searchable indices. The encrypted images along with search indices are outsourced to cloud. The image owner computes the index for query image and retrieves top k images near to similarity and decrypts the image with key shared by image owner. Experimentation was performed on Corel-10K dataset for five major categories and precision obtained was satisfactory and the privacy of image content also preserved.