Prediction of beef CIELAB colour from RGB digital images

The aim of this study was to evaluate the performances of a quadratic model to predict CIELAB parameters from digital images. The colour of 80 longissimus thoracis samples was measured by a spectrophotometer in the CIELAB space model and photographed using a digital camera which produced RGB images. All the images were captured under controlled conditions. The RGB colours were measured using Photoshop software. The conversion of RGB values to L*a*b* values was carried out using a quadratic model. The percent mean absolute error (e ̅ %,), standard deviation of the percent mean absolute error (), average root mean square error ((RMSE) ̅) and colour difference (ΔE*) were used for measurement of differences between the values obtained with the spectrocolorimeter and Photoshop. The model showed an error of 1.36% and a standard deviation of 1.12. The ((RMSE) ̅) was 1.28 while the ΔE* was equal to 2.94. The proposed method achieves a promising performance, however the acquisition of the images needs some adjustments to improve the accuracy of the model.