Efficient Error-Tolerability Testing on Image Processing Circuits Based on Equivalent Error Rate Transformation

Abstract

In this paper we address two key issues related to error-tolerability testing on image processing circuits, namely acceptable threshold determination and acceptability evaluation. A JPEG 2000 decoder is employed as a case study. We first carefully investigate the acceptability threshold values of images in terms of error rate and error significance. The investigation results show that appropriate threshold values should be determined depending on a number of factors, including size (resolution), brightness, contrast and frequency of test images as well as human subjectiveness. Based on the determined threshold values we propose an equivalent error rate transformation technique to help test engineers easily and quickly examine the acceptability of a circuit under test. We also carry out hardware implementation of the proposed technique. The implementation results show that the hardware area overhead with respect to a commercial JPEG decoder design is only 2.24 %. To validate the proposed technique, 13,860 erroneous color images produced by faulty JPEG 2000 decoders as well as 450 erroneous benchmark images are employed. The experimental results show that the proposed technique is as effective as the exhaustive test method. Moreover, our technique requires much less test time and storage space compared with the exhaustive test method. The achievable reduction ratio in test time and storage space is more than 99 %.