White Papers Home
Understanding Frame Rate
This document provides an introduction, to all level of readers, the effect of frame rate selection on video compression, with specific reference to stills-based codecs, such as MJPEG, and motion-based codecs, such as MPEG-4. The document demonstrates that even at low frame rates that motion-based encoders will surpass the performance of stills-based coding. Further, that the cost of increasing the frame rate with motion-based encoders is significantly less than that of still-based counterparts.
Contents
2 EFFECT OF ALTERING FRAME RATE
1 Background
This document explores the effect of frame rate selection within a digital video encoder, with specific reference to the effect on video quality and bitrate. Through a series of experiments, using the IndigoVision 8000 series encoder, the reader should come realise that increasing the frame rate of a motion-based encoder, such as MPEG-4, is far less costly than similarly increasing the frame rate of a stills-based encoder, such as MJPEG. Further, that even at low frame rates a motion-based encoder will surpass the performance of a stills-based encoder in terms of video quality and bitrate.
2 Effect of Altering Frame Rate
The natural full frame rate of an interlaced camera is fixed at 25fps for PAL cameras and 30fps for NTSC. Encoders, such as the IndigoVision 8000 MPEG-4 motion-based encoder, can compress full frame rate video through a process of removing the unnecessary information that exists within each individual frame of video, but that also exists over periods of time, such as a static background.
Other types of encoder, such as an MJPEG stills-based encoder, remove only the unnecessary information within each individual frame of video, regardless of the information that has already been transmitted by previously compressed frames.
Altering the frame rate will affect both types of encoder. In the next couple of sections two experiments are run. In the first a CCTV video sequence is used, which has some motion but also a lot of static information. In the second experiment a far more active sequence is used with plenty of fast moving action. Both sequences are encoded at a series of different frame rates using a standard motion-based encoder and a standard stills-based encoder.
2.1 CCTV sequence
The CCTV sequence used is an office camera, with some limited motion. A frame of the sequence is shown in Figure 1.
Figure 1: CCTV scene
This sequence is encoded, as previously stated, with a motion-based encoder and a stills-based encoder at a selection of different frame rates. The bitstreams are encoded with exactly the same video quality, and the resulting bitrate measured. The result of the experiment is shown in Figure 2.
Figure 2: Frame rate v bitrate for same quality CCTV
Figure 2 shows that as the frame rate increases, the bitrate increases for both encoders. At the lower frame rates the difference is the order of 100kbps. However, as the frame rate increases the motion-based encoder always produces increasingly fewer bits for the same quality of video, than the stills-based encoder, to the point that at 30fps the difference is of the order of 1.5Mbps!
This can be explained simply as the static background parts of the video remain static for long periods of time. This means that, even at the lower frame rates, the motion-based encoder can work with only a slight reduction in efficiency.
What about when there is less static background video?
2.2 Movie sequence
The experiment is repeated with a new sequence. In this sequence there is much more complicated, high motion, high action video. An example frame from the sequence is shown in Figure 3.
Figure 3: Movie scene
The results for this sequence are shown in Figure 4. The results show that for this more complicated motion sequence that the motion-based encoder once again produces fewer bits for the same quality video as the stills-based version. However, understandably the difference is less marked in this case.
Figure 4: Frame rate v bitrate for same quality movie
3 Conclusion
Motion-based encoders, like MPEG-4, will outperform stills-based encoders, like MJPEG, in terms of either video quality or bitrate, as they always have an option to make use of any similarities that exist over time, be it static background or coincidentally matching data.
Reducing the frame rate simply reduces the probability of the motion-based encoders finding similarities and thus being less efficient. However, it will always be better than not trying at all, as with stills-based encoders.
Conversely, increasing the frame rate increases the probability of the motion-based encoders finding similarities and thus being more efficient. In a CCTV scenario it has been shown that increasing the frame can lead to a 11-fold increase in bitrate with a stills-based encoder, when compared to a less than 5-fold increase with a motion-based encoder. Furthermore, that even with very complex, high motion video not found in CCTV applications, that the relationship, if not the degree, remains true.
