In order to explore the feasibility of hyperspectral imaging technique to estimate beef color parameters under different storage time and sampling positions, hyperspectral images of 82 representative beef samples were acquired. Color parameters, including brightness (L*), redness (a*), yellowness (b*) and saturation (C*) were also determined. Their representative spectra were obtained by selecting regions of interest (ROIs). By comparing and choosing appropriate spectral regions and pretreatment methods, optimum partial least squares (PLS) calibration models of each beef color parameters were established and evaluated, respectively. As for L*, a*, b* and C*, the correlation coefficients of calibration were 0.80, 0.91, 0.91 and 0.93, and root mean square errors of calibration were 2.23, 1.18, 0.82 and 1.12, respectively. The correlation coefficients of prediction were 0.92, 0.88, 0.87 and 0.89, and root mean square errors of prediction were 1.66, 1.45, 0.80 and 1.27, respectively. The results showed that hyperspectral imaging technique could be used to rapidly and non-destructively analyze beef color parameters under different storage time and sampling positions.