Theoretical and computational studies have suggested that the visual cortex processes natural sensory information with characterized pattern that is termed as "sparse coding", which means that each individual neuron rarely fires intensely (lifetime sparseness), and meanwhile, only a small subset of neurons within a large population are activated in response to a given instantaneous stimulus (population sparseness). Temporal and spatial patterns of the chicken retinal ganglion cells′(RGCs) activities in response to time-varying natural images (movies) as well as pseudorandom white-noise checker-board flickering sequence (control) were analyzed. The sparseness indices of the RGCs′ response over lifetime and across population were calculated, the detailed temporal and spatial characteristics underlying such sparseness were also investigated. The results show that the lifetime sparseness and the population sparseness were both more profound for the neuronal responses evoked by natural stimuli as compared to that elicited by checker-board flickering. Further analysis shows that there were more action potentials fired in "burst" form in response to natural stimuli. Coincident bursts of adjacent neurons were prevalent in response to both kinds of stimulation, but occurred more frequently during natural movies stimulation. These results suggest that the RGCs encode natural sensory inputs efficiently. In this scheme, individual neuron fires at a low rate to save metabolic energy, while dynamically grouped small subsets of neurons are activated with adjacent neurons firing concertedly to transmit information to the postsynaptic neurons efficiently.