ABSTRACT: The average colors of mangos and apples were measured using machine vision. A method to quantify the perception of nonhomogeneous colors by sensory panelists was developed. Three colors out of several reference colors and their perceived percentage of the total sample area were selected by untrained panelists. Differences between the average colors perceived by panelists and those from the machine vision were reported as ΔE values (color difference error). Effects of nonhomogeneity of color, and using real samples or their images in the sensory panels on ΔE were evaluated. In general, samples with more nonuniform colors had higher ΔE values, suggesting that panelists had more difficulty in evaluating more nonhomogeneous colors. There was no significant difference in ΔE values between the real fruits and their screen image, therefore images can be used to evaluate color instead of the real samples.