This paper presents a practical evaluation of a RADIa1 Soft Hash algorithm, an image hashing technique used for image and visual content authentication. First we evaluate its collision resistance and susceptibility to geometrical deformations and image processing attacks. Second we investigate its actual efficiency on different platforms. Although software implementations allow managing sizes and throughput for most applications, they suffer from some troubles for real time applications such as video and for images in high definition. For these more constrained contexts, we propose hardware implementations. In practice, throughput up to 2^{18} (resp. 2^{26}) pixels/sec are managed in a single SW (resp. HW) implementation.