Sediment movement in mountain areas is often induced by heavy rainfall. Hence, rainfall observation in mountain areas is crucially important for prevention of sediment‐related disasters. Although rainfall is measured with tipping bucket rain gauges and weather radar all over Japan, we have a technical challenge in respect of spatio‐temporal resolution for the observation network. Therefore, we attempt to develop a novel method for estimating rainfall intensity at a high spatio‐temporal resolution with closed‐circuit television (CCTV) surveillance cameras. We collected videos of three CCTV cameras during rainfall events and acquired digital image data from those videos at the one‐minute interval. Previous analyses showed that the contrast and color of images are drastically affected or degraded under rainfall conditions. Based on the results of these analyses, we analyzed the relationships between rainfall intensity and pixel values. Results show that the pixel values go higher and frequency distributions of pixel values become narrower with increasing rainfall. Moreover, the introduced pixel value indices which are means of pixel values in analytical patches on images fluctuate in response to rainfall in time series. The results indicate that the indices are applicable to observe the change of rainfall intensity. We also showed that place and size of analytical patches on images is important for this analysis. Based on these results, we estimated fluctuations of the rainfall intensity from the fluctuation of the index at one‐minute interval. Consequently, we will be able to better estimate amounts of rainfall at a high spatio‐temporal resolution in mountain area by utilizing the current operational CCTV camera network.