Cyclic variations constitute an inherent consequence of the flow, thermal and concentration field variations between cycles. They are understood to lead to lower efficiency and higher emissions. The current investigation aims to evaluate the cycle-to-cycle variations (CCVs) based on 2D visualization and cylinder pressure in an optically accessible heavy-duty engine fueled with methane (main fuel) and diesel (pilot fuel). A high-speed color camera is employed to measure the combustion behavior based on natural luminosity (NL). Proper orthogonal decomposition (POD) is applied to reconstruct and analyze the images. The POD-based coefficient of variation (COV) is implemented to evaluate the cyclic variability, along with the pressure-based and global intensity-based COV. This coefficient is then adopted to discriminate the coherent and incoherent parts from the fluctuations in the luminosity field. The POD-based and global intensity-based COV presents the variations in the luminosity field, which can provide information on chemical kinetics, while pressure-based COV provides a general description of the cyclic fluctuation of thermodynamics. To extract more information from the NL images, the color-intensity COV analysis based on the intensity separated from RGB channels is adopted to estimate the CCVs from the aspect of spectral emissions (excited and ionized radicals in the flame). Finally, the effects of methane lambda, pilot fuel rate and charge air temperature on the CCVs were analyzed systematically. The results revealed that richer methane conditions has an inhibitive effect on the CCVs. The appearance of the CCVs were determined by the ignition characteristics of the pilot fuel. A critical point was found in charge air temperature, when the charge air temperature lower than the critical point, the increase of the charge air temperature has a promotive effect on the CCVs; after that, it has an inhibitive effect on the CCVs.