文摘
We propose a robust feature extraction method for color images, which is then exploited to design a perceptual image hashing algorithm. The feature extraction method is achieved by converting the RGB color image into HSI and YCbCr color spaces and extracting the block mean and variance from each component of the HSI and YCbCr representations. In hash generation, we re-scale the image to a fixed size, calculate the color features by the proposed feature extraction method, and take the Euclidian distances between block features and a reference feature as hash values. Experimental results indicate that the proposed hashing algorithm is robust against the content-preserving manipulations, such as JPEG compression, brightness adjustment, scaling, and small angle rotation. It shows good performances in discriminative capability, and can detect malicious tamper in the local image areas.