Noisy video is a big headache for video compressionists. Random noise causes pixels to change rapidly and constantly, making it difficult for the codec to encode the clip correctly.
Noise reduction encompasses a variety of techniques to remove that noise from an image, making it easier to compress. Very simple noise reduction algorithms are just blurs that hide grain but make the image softer as well, such as when a median filter is applied (Figure 4.5). More advanced algorithms try to blur only the parts of the image that have grain and may take advantage of differences among frames.
Figure 4.5 Here are two versions of a DV-NTSC clip. On the right it has had a median filter applied. Note the increased softness in the image on the left but the smoother quality of the image also.
Noise reduction is not always required. If you have a clean video source to work with, simply skip the noise reduction step when preprocessing your work. If you do have damaged footage, it may take several experiments to find the right balance of cleaning up the source without degrading the finished video image.