A measure for assessing edge fidelity of coded images based on shape analysis.

The authors propose a method for quantifying the edge degradation of a coded image produced by aliasing distortion or block distortion based on the change in its shape. Essentially, when a change is generated in a smooth portion, a human can quickly detect it. With conventional techniques based on signal differences, it was difficult to assess a geometric distortion for which continuity or uniformity changed. The proposed technique obtains mutually corresponding edge segments for the original image and coded image and quantifies the shape degradation based on their curvature. To match subjective assessments, the measure that was obtained takes into consideration (1) the spatial size of the shape change, (2) the shape error masking effect, and (3) the edge gradient (contrast) effect in the quantification process. The authors performed simulations for several standard images while changing the coding rate and coding method, and the results that were obtained reflected the impressions of the images well. © 2004 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 87(8): 9–23, 2004; Published online in Wiley InterScience (<URL>www.interscience.wiley.com</URL>). DOI 10.1002/ecjc.10157 [ABSTRACT FROM AUTHOR]