Aged films contain valuable historical information. However, some films may have defects due to dirt, scratches or for other reasons. Usually, there is no efficient prediction model to precisely identify these defects. This study proposes a set of new algorithms based on the analysis of features among frames in aged films. Both spatial and temporal characteristics are used. The algorithms further use filters with adjustable thresholds to precisely identify defects, such as spikes and scratches. Finally, a temporal inpainting method and a multiple resolution image inpainting algorithm are used to repair the defects. Since no efficient automatic quantitative mechanism is available due to the loss of original films, the experiments use subjective evaluation on more than ten films. The results show that most defects are repaired with a pleasant visual quality.