Histogram-based image hashing scheme robust against geometric deformations

Abstract

In this paper, we propose a robust image hash algorithm by using the invariance of the image histogram shape to geometric deformations. Robustness and uniqueness of the proposed hash function are investigated in detail by representing the histogram shape as the relative relations in the number of pixels among groups of two different bins. It is found from extensive testing that the histogram-based hash function has a satisfactory performance to various geometric deformations, and is also robust to most common signal processing operations thanks to the use of Gaussian kernel low-pass filter in the preprocessing phase.

A Survey on Image Hashing for Image Authentication

“…The common preprocessing operations include image downsampling [7], low-pass filtering for reducing high frequency signals [8], resizing for image rescaling, order statistic filtering for denoising [9], and Gaussian blurring [10]. As such, all above operations serve for the next essential attributes extraction stage.…”

“…Hence, the feature extraction stage is significantly important in order to achieve robustness, discrimination and security. Some of the robust feature extraction methods exploited in literature include image histogram [10], feature points [11] or image edge information, significant DWT or DCT coefficients, Fourier transform [8], and dimensionality reduction with linear transforms [12].…”

“…The clustering techniques [1], [13], decoding step of a proper Error Correcting Code (ECC) [9] have been exploited for compacting an intermediate hash. Furthermore, a key-based permutation can be applied to hash vectors for randomization [10]. However, the security is not as strong as in previous feature extraction stage, also the accuracy will be varied with the changes of key inserting [12].…”

“…The two images are regarded as perceptually same if the distance is close to 0, whereas the distance is expected to be close to 0.5 for two perceptually different images. Due to its simplicity and efficiency, the normalized Hamming distance is the most widely used measurement metric in image hashing systems, especially for those with the postprocessing containing one-bit quantization or relationship-based bit generation, such as the approaches in [10], [16].…”

An Improved Hashing Method for the Detection of Image Forgery

“…In a robust histogram based image hashing scheme, the robustness of the hash against geometric deformations is achieved by using the histogram shape invariance. Real-world image authentication is not possible with this approach [3].…”