IJCSES International Journal of Computer Sciences and Engineering Systems CSES International ⓒ2008 ISSN 0973-4406 1 Manuscript received May 6, 2008. .                                                Chung-Hao Chen, Yi Yao, David Page, Besma Abidi, Andreas Koschan, and Mongi Abidi Imaging, Robotics, and Intelligent Systems Laboratory Department of Electrical Engineering and Computer Science The University of Tennessee Knoxville, TN 37996 USA E-mail: {cchen10, yyao1, dpage, besma, akoschan, and abidi}@utk.edu Abstract In this paper, we provide objective measures to evaluate compression methods for machine recognition applications. Vidware Vision, a black box compression method developed by Vidware Incorporated, is used in the case and its performance is compared to existing compression methods, i.e., JPEG and JPEG 2000, based on various measures. The encoding and decoding time are used to characterize computational complexity. Full- and no- reference image quality measures are exploited to describe distortions and degradations in the decompressed images. In addition, since this paper focuses on the performance of compression methods relating to machine recognition applications, we propose a non-separable rational function based Tenengrad (NSRT 2 ) measure to evaluate the sharpness of decompressed images. Based on our experimental results, Vidware Vision TM is robust to changes in compression ratio and presents gradually degraded performance at a considerably slower speed in terms of computational complexity and image quality. Particularly, according to full-reference measures Vidware Vision outperforms JPEG and JPEG 2000 when the compression ratio is larger than 140. The effectiveness of our proposed NSRT 2 , as a new comparison tool, is also validated via experiments and performance comparisons with other tested measures. . Key words: JPEG, JPEG 2000, Vidware Vision, Image Quality Measure, Image Sharpness Measure. 1. Introduction Image compression [1, 2] is essential for image storage [5], transfer [7], access [5], and other related applications [4, 6, 8, 10]. The goal [4, 6, 10] of compression is to minimize the size of the data being broadcasted or stored, thus minimizing transmission time and storage space, while maintaining a desired quality compared to the original image. One of the most popular lossy compression approaches for still images is JPEG. JPEG [3, 9] stands for Joint Photographic Experts Group, the name of the committee that developed the standard. JPEG compression eliminates unnecessary data, according to the characteristics of the human eye, resulting in significantly reduced file size at the cost of slight degradation in image quality. JPEG 2000 [2, 8, 14], a wavelet-based image compression standard, was also created by the Joint Photographic Experts Group with intention to outperform their original JPEG standard which uses an 8×8 block-size discrete cosine transform. Compared with JPEG, JPEG 2000 is not only superior to JPEG in terms of subjective image quality, but it provides additional functionalities such as: lossless and lossy compression, spatial and quality scalabilities, tiling, region of interest coding, random bit stream, error resilience, content-based description, and protective image security. Vidware Incorporated developed a product line called Vidware Vision [12, 13], which is a black box compression method for end users. It is divided into three separate categories: Still Image (as a replacement for JPEG), Full Frames video (as a replacement for M-JPEG), and Full Motion video (an H.264 compliant CODEC). The fundamental key of image compression is to reduce data redundancy. Certain degree of data loss is acceptable as long as it is not noticeable to the viewer, where the “viewer” is not necessarily a human. The viewer can also be an image processing algorithm. Thus, the performance evaluation of various compression methods and the choice of evaluation measures also depend on the targeted “viewer”. In addition, theories behind compression methods are not always available or understandable for end users in most cases. Thus, evaluation measures suitable for “black box” compression methods ought to be selected according to following two aspects: (1) computational complexity and (2) image quality.