All panelists were screened for formal train before experiments. Every panelist was re-familiarized with the attributes of beef-marbling image in sessions where panelists were retrained to determine the scores of an algorithm performance. Training sessions were terminated when all panelists were ±1/6 of a score from allover mean score and not more than one unit from the group mean score specified for each sample. Note that the geometric mean of five scores, retaining a decimal point after the effective figures, is used to define the score of a segmentation algorithm performance.