Identification of Critical Errors in Imaging Applications

Practical on-line test methods do not cover all possible faults of a system. We propose a method to identify critical faults and distinguish them from non-critical ones. Low-cost on-line fault detection can focus on the critical faults. Alternatively, the circuit sites associated with critical faults could be selectively hardened to improve the overall reliability of a system. This is done in a cost- effective way because no hardening against non-critical faults is required. In this work, we concentrate on faults in imaging applications such as video. We classify faults based on their impact on the system behavior, i.e., the visibility of their effects by a human end-user. The psycho- visual model from the JPEG compression method is used for fault effect classification.