The apple CT gray scale images scanned by CT were transformed by Fourier transform, and 16 parameters were extracted from each frequency domain after transformation. Combined with the soluble solid, the titrable acidity, the pH value and the moisture content of apple, the principal components regression (PCR) and the partial least squares regression (PLSR) were employed to establish the prediction models of apples’ internal quality. In the PCR, the first ten principal components were chosen with contribution rate reaching 99%. The models show good prediction results by the F criterion with all the P values lower than 0.05. In the PLSR, each content model has the lowest sum of squared errors when the number of latent variables is 12, which indicated a good prediction result. The results show that the models built by PCR have higher predictive ability than that of PLS method in the matter of errors.