A Matlab Tool for Subjective Assessment of Image Quality

Image quality assessment (IQA) is to evaluate the performance of image processing techniques, and acts as an important way to promote their further developments. In objective evaluation, the image quality of a to-be-evaluated image is related to its difference with the ground-truth image which is estimated by using various metrics, such as the average error and the mean square error. However, an objective evaluation is often impossible to exactly match the perception of human vision system. Therefore, a reliable subjective assessment of image quality is always necessary. In this paper, a Matlab tool is designed to facilitate this evaluation task from the following four aspects. First, every to-be-evaluated image is presented to viewers with a random order and without any label; that is, a blind evaluation is conducted. Second, after a round of evaluation, the order of the to-be-evaluated images will be randomly changed and a second round of blind evaluation will be required; that is, a double-confirmed evaluation is achieved. Third, during the evaluation, the current image (either the ground truth or a to-be-compared one) can be enlarged by the viewers for a close watching. In particular, when this image is enlarged, other images in the same group will be automatically enlarged at the same portion for fascinating the comparison; that is, a synchronous magnification is realized. The final evaluation result is taken to be a mean of the scores of the two evaluation rounds, provided that their difference is below a given threshold; that is, a automatic analysis is achieved. The tool is publicly available at https://github.com/hdddx/SubjectiveEvaluation.