Effect of the degree of non-homogeneity of color on the sensory panel color quantification error

Institute of Food Technologists (IFT) Annual Meeting & Food Expo, Las Vegas, United States Of America, 25 - 28 June 2012, pp.252, (Full Text)

1. Justification

Color is an important quality attribute and one of the most attractive features of the Salmonidae

family of fish. The color of the whole fish is one indication of its quality for consumers.

However fish color is non-homogeneous. Human perception of non-homogeneous color is not

well understood.

2. Objectives

Our objective was to determine of effect of the degree of non-homogeneity on the error of the

overall color quantification of some Salmonidae fish by a sensory panel.

3. Methods

Four Salmonidae fish species (Salmo trutta lacustris, Salmo trutta macrostigma, Salmo trutta

forma fario, Oncorhynchus mykiss) images were first analyzed to calculate average L*, a* and

b* values. Then color primitives of increasing “averaging” were applied to reduce nonhomogeneity

while preserving overall color. 69 university students attended the sensory panels.

4 reference colors out of 10 were selected by the panelists and their % areas in the image were

estimated. The difference between the color predicted by the panelist and the real average color

was expressed as ?E error.

4. Results

There was no significant difference between male and female panelists in their prediction of the

average colors. The reduction of the degree of non-homogeneity by local color averaging (color

primitives) did not reduce the error made by the panelists in estimating colors. Repeating the

panel did not reduce the error. When the “best” color combinations and their percentages were

calculated by the computer, it was evident that the given color references could very accurately

predict the average color of the fish (?E < 0.1).

5. Significance

The results of this study suggest that humans have difficulty in estimating the average color of

samples with non-homogeneous colors. Machine vision, on the other hand can quantify the

average colors easily, and should be used for this purpose.