Reducing the duration of paired-comparison experiments for visual quality assessment

The subjective methodology of paired comparison is currently recognized as the most precise methodology. However, it is not widely used because it implies a larger number of assessments than other methodologies and therefore a longer test duration. In this paper we investigate how sorting algorithms can be used to decrease the duration of such experiments by selecting only a fraction of all the possible comparisons. Three methods based on the insertion sort and the binary tree algorithms are applied to paired-comparison data. Performance analysis shows interesting trade-offs between the number of comparisons and the accuracy of the results.