A fingerprinting is related to cryptographic hash functions. In contrast to cryptographic hash functions this robust digest is sensitive only to perceptual change. Minor changes, which are not affecting the perception, do not result in a different fingerprint. This technique is used in content-based retrieval, content monitoring, and content filtering. In this paper we present a cumulant-based image fingerprinting method. Cumulants are typically used in signal processing and image processing, e.g. for blind source separation or Independent Component Analysis (ICA). From an image with reduced dimensions we calculate cumulants as an initial feature vector. This feature vector is transformed into an image fingerprint. The theoretical advantages of cumulants are verified in experiments evaluating robustness (e.g. against operations like lossy compression, scaling and cropping) and discriminability. The results show an improved performance our method in comparison to existing methods.