Perceptual Image Hashing: Tolerant to Brightness and Contrast Corrections Method Based on Cumulative Histogram Slicing

Abstract

Perceptual image hashing is used in a wide range of practical applications which include content image authentica- tion, digital watermarking, pattern recognition, computer vision and database fast duplicate image retrieval. Existing techniques are not well suited for the significant brightness and contrast corrections. The main point is that such manipulations can lead to information loss due to the histogram truncation in cases when pixel values are out of the dynamic range. In order to address the issue a novel technique is suggested. Cumulative histogram slices as a pivot for the subsequent image features calculations are used. The points of slicing are calculated in a way they are robust to content preserving manipulations such as brightness and contrast corrections. This approach allows one to handle situations when some of the content slices are lost due to the pixel value overflow. On the other hand, if one tampers image content within any existing slice it will then be detected by comparing the correspondent calculated and provided hash values. Experiment results show that the suggested method has sufficient sensitivity to detect image tampering whereas being tolerant to even significant brightness and contrast corrections. The memory consumption allows one to use the proposed method with the digital watermarking schemes.

Keywords

• perceptual image hashing
• digital watermarking
• content image authentication
• cumulative histogram slicing
• brightness
• contrast