An automated grading method of chicken weight using image processing was proposed. Ninetyfive images of chicken were acquired randomly in a poultry slaughtering plant by using a digital camera. After these images were preprocessed, six parameters such as projection area (Sp), contour length (Cp), length (Hp), breast width (Ap), breast length (Bp) and fitting ellipse (Ep) of chicken carcass were extracted from the processed images. Then taking the six parameters as the inputs and ninety five samples as the training set， the simple linear regression model and multiple linear regression model were established for predicting of chicken weight, respectively. Furthermore, the optimal model was found out among these developed ones according to regression correlation coefficient. Finally, the independent validation set was formed by using 100 samples divided into five groups and employed to validate the optimal model. Results showed that the simple linear model based on the projection area (Sp) of the chicken carcass had the largest R2 of 0.827 in the six simple linear models developed. The multiple linear regression model developed based on the indicators of Sp, Cp, Ap and Bp had the largest R2 of 0.880 in all multiple linear models developed. The adjusted multiple linear regression model had a adjusted R2 of 0.933 after eliminating eight outliers detected by students residuals. When the validation set samples were used to validate the optimal multiple linear model, the average correct rate for weight grading of chicken carcass was 89%, indicating that the proposed method based on image processing was feasible for automatic weight grading of chicken carcasses.