A Texture Based Image Signature Using Second Order Statistics Characterisation
This paper develops a scheme for semi-fragile image authentication based on texture features and digital signature. It can detect and locate malicious manipulations made to individual image blocks, and verify the integrity of the overall image. It makes u
- PDF / 155,380 Bytes
- 2 Pages / 430 x 660 pts Page_size
- 105 Downloads / 219 Views
Abstract. This paper develops a scheme for semi-fragile image authentication based on texture features and digital signature. It can detect and locate malicious manipulations made to individual image blocks, and verify the integrity of the overall image. It makes use of the invariance of Second Order Statistics (SOS) on adjacent pixels to geometric transformations which are considered as content preserving modifications. A set of characteristics are extracted from each block, then processed to form a hash to be embedded in an other distant block. Furthermore, a cryptographic digital signature is incrementally formed, in a CBC manner that renders the scheme resistant to content cutting and pasting attacks.
1 Features Based Image Authentication The proposed features detection scheme is especially sensitive to texture alterations, while being invariant with respect to geometric transformations. The most accurate techniques for analysing image texture are statistical methods which analyse the spatial distribution of grey values, by computing local features and deriving a set of statistics from the distributions of local features. Depending on the number of pixels defining the local feature, the statistical methods can be respectively classified into first-order, second-order and higher-order statistics. Haralick [1] suggested the use of grey level co-occurrence matrices (GLCM) to extract second order statistics from an image. The joint probability of grey levels for two pixels is calculated with respect to a distance d and an angle θ, then stored in a matrix which can be used to extract 14 different second-order statistical texture features, describing the probability density function. GLCMs have been used very successfully for texture classification, segmentation and content based image retrieval CBIR [2]. The specific features considered in this work are: N −1 N −1
2
N −1 N −1
Inertia = ∑∑ (i − j ) p (i, j ) , Dissimilarity = ∑∑ i − j . p(i, j ) i =0 j = 0
i = 0 j =0
N −1 N −1
N −1 N −1
i =0 j =0
i =0
Entropy = ∑∑ p (i, j ) log p (i, j ) , Homogeneity = ∑∑
1 p (i , j ) 2 j = 0 1 + (i − j )
We extract texture information in a particular way: From each 8X8 pixels block, GLCM are constructed at a distance of d = 1, 3, 5 and 7 and at angles 0°, 45°, 90°, R. Meersman, Z. Tari, P. Herrero et al. (Eds.): OTM 2007 Ws, Part I, LNCS 4805, pp. 44–45, 2007. © Springer-Verlag Berlin Heidelberg 2007
A Texture Based Image Signature Using Second Order Statistics Characterisation
45
135°. Each of the obtained co-occurrence matrix is weighted on a spatial position basis value then averaged. To make the features invariant towards rotation, the obtained matrices are summed over the four angles to give a single matrix from which the 4 aforementioned statistics are extracted giving an approximation of the block’s textural property. The chosen method for embedding the extracted features is an adapted version of Wu & Tsai data hiding technique [3]. In addition of being a high capacity low computational cost technique, it is
Data Loading...
