Disclosed are methods, devices, and computer program products for image noise reduction. In one example embodiment, a method for reducing noise in a digital image includes several acts. First, one or more objects are identified in an input image. Next, the input image or a portion thereof is processed to produce another image, which may be a blurred version of the input image. Finally, one or more objects in the processed image that correspond to the one or more objects identified in the input image are evaluated to determine whether to discard the one or more objects identified in the input image. For example, if an amount of energy preserved in an object after blurring is less than a threshold, the object may be discarded as noise.