Quality-Based Transcoding (QBT) – Simplifying Transcoders For Target Video Quality

For video streaming applications, three significant requirements are (1) high video quality, (2) a reliable system, and (3) optimal cost of the offering.

The current solutions in the market offer video transcoding with user settings for different transcoding parameters. It is difficult for media content companies to choose appropriate transcoding settings because –

• multiple encoding parameters require expert knowledge,
• choosing appropriate settings for different types of video content is not practical.

Ashok Magadum is the Founder and CEO of Vidarka Technologies; a company focused on the domain of High Quality, Cost Optimal, Scalable, and Reliable Video Streaming Solutions for VOD, Live and Low Latency Streaming covering OTT/FAST, EdTech, E-Shopping, Gaming, and many more application areas.

Quality-Based Transcoding (QBT) offers simple target quality settings for transcoding to media companies so that they can achieve the target video quality with optimal bit rates and, in turn, achieve an optimal cost.

In this article, we will share the results of Quality Based Transcoding for 1080p H.264/AVC and 720p H.264/AVC for mobile devices.

Also, do check out my earlier articles on H.264/AVC vs. HEVC and Chunk-based High-Speed Transcoding.

Quality Mode and Target VMAF for the Video transcoding for different target devices

Target Quality for 1080P H264 Video Encode Mode
Mode	Target device	Target VMAF	Content
Mobile	Mobile and Tablet	95 (Mobile)	Edtech, Sports, OTT, E-Shopping
Economical	TV	88	OTT: TV Shows, Movies, Sports
Standard	Bigger size screen TV	92	Standard or High-Quality Movies
Premium	High-Quality Movies, Sports and Shows	95	Premium Movies, sports and shows
HDR	HDR TVs	97	HDR Content

Need for Target Quality-based transcoding

Target quality-based transcoding refers to adjusting a video’s quality during transcoding to meet a specific target quality level.

There are several reasons why target quality-based transcoding is essential:

Bandwidth and Storage

Videos with higher quality require more bandwidth (CDN) and storage space. If the target quality level is too high, the video may take longer to upload or download and may take up too much storage space. Adjusting the target quality level allows you to optimize the video for its intended purpose and save on storage and bandwidth costs.

User Experience

A video that is too low in quality may be challenging to watch or understand. At the same time, a video that is too high in quality may take too long to buffer or may be incompatible with specific devices. By optimizing the video for its target quality level, you can ensure that it provides the best possible user experience.

Quality Control

Target quality-based transcoding can also help ensure the video meets specific quality standards. For example, if you are creating a professional video, you may need to ensure that it meets specific quality criteria; by setting a target quality level, you can ensure that the video meets these requirements.

Conventional Transcoding settings

Popular video transcoding engines provide support for various video formats.

The most popular video compression format is H.264/AVC.

Relatively newer formats HEVC, VP9, AV1, and VVC, are slowly gaining popularity. The settings which modify, fine-tune, and improve video coding efficiency are bitrate, IDR period/Key interval, GOP structure, Adaptive Quantization, scaling, and many more.

Challenges in choosing the options

The video quality is directly affected by several factors, and some of these are –

Key Frame interval and GOP structure.

• Longer GOP structures can lead to higher compression ratios, resulting in smaller file sizes but lower video quality in erroneous conditions.
• In contrast, shorter GOP structures can improve video quality but result in larger file sizes.

Complexity of the Content:

The complexity of the video content can affect the choice of encoding parameters. For example, videos with fast-moving scenes or complex textures may require shorter GOP structures to maintain high-quality video.

Available Bandwidth

The available bandwidth can also affect the choice of encoding parameters. Lower bandwidth may require more aggressive compression techniques, such as longer GOP structures or higher compression ratios. In comparison, higher bandwidth may allow for more relaxed parameters that result in higher-quality video.

Time and Computational Resources

Encoding video can be a computationally intensive process, and choosing the optimal encoding parameters may require significant time and resources to experiment with different settings and compare results.

How Target Quality-based Transcoding helps

When the encoding decision is left to the encoder based on the target quality and provided bitrate range, it can pick appropriate settings for various scenes and encode them, achieving the best quality and saving the bitrate significantly.

Two-pass encoders can achieve this by analyzing the video in the 1st Pass and encoding the video in the 2nd pass. But, it has certain limitations –

Two Pass Encoder Limitations

A two-pass encoder is a video compression technique that analyzes the video content in two passes to determine the optimal encoding parameters for the highest quality and compression efficiency. While two-pass encoding can provide superior results to single-pass encoding, it has some limitations.

Here are a few:

Time-consuming

Two-pass encoding can be more time-consuming than single-pass encoding, requiring two separate encoding passes. This can be a limitation for real-time applications or situations requiring fast encoding.

Less Flexible

Two-pass encoding requires that the video content be fully available before encoding begins. This can be a limitation for live streaming or other real-time applications where the video content is generated in real-time.

Increased Complexity

Two-pass encoding is a more complex technique than single-pass encoding, requiring additional processing and analysis of the video content.

Single Pass Encoder and Advantages

Single-pass video encoding, or one-pass encoding, is a video compression technique where the video is encoded in a single pass without needing multiple analysis passes. Here are some advantages of single-pass video encoding:

Speed

Single-pass encoding is generally faster than multi-pass encoding methods because it avoids the additional time required for multiple analysis passes. This makes it suitable for real-time encoding or situations where time is a critical factor.

Simplicity

Single-pass encoding is simpler to implement and use since it eliminates the need for complex analysis passes and the associated decision-making processes. It can be an efficient choice for applications that require straightforward and quick video encoding.

Real-time streaming

Single-pass encoding is particularly useful for real-time streaming applications like video conferencing, live broadcasting, or online gaming. It enables the encoding process to keep up with the real-time input, ensuring minimal delay and smooth playback.

Reduced computational resources

By eliminating the need for multiple analysis passes, single-pass encoding reduces the computational resources required. This can be advantageous in scenarios with limited hardware or computing power, such as on mobile devices or low-end computers.

Single pass encoder with few frames look ahead can provide the best prediction for Quality-Based Transcoding. Low complexity and faster turnaround time help to use this encoder for live video transcoding use cases.

H.264/AVC Encoder Quality Settings

VMAF (Video Multimethod Assessment Fusion) is a popular video quality metric that measures the perceived quality of a video. The optimal target VMAF (or VMAF score) for different video content can vary depending on the intended use, delivery platform, and viewer preferences. Here are some general guidelines for target VMAF scores for different types of video content:

Streaming platforms and Online video

For most streaming platforms and online video services, a target VMAF score between 86 and 92 is often considered a good quality. This range balances efficient compression and maintains a visually pleasing viewing experience.

High-definition (HD) content

HD content typically requires higher-quality targets. A target VMAF score of 88+ is desirable for good quality video, and a VMAF of 92 or above is often desirable for HD videos to ensure a clear and crisp viewing experience, especially for content with intricate details or fast motion. Higher VMAF provides ensures better quality.

HDR video designed to provide more color gamut and luminance range needs lesser compression losses to preserve the best quality.

Virtual Reality (VR), 360-degree videos, and Metaverse streaming

VR and 360-degree videos require a high level of immersion and quality to deliver an engaging experience. A target VMAF score of 92 or above is generally recommended to minimize artifacts, blurriness, or other issues that can diminish the sense of presence in VR environments.

Mobile Mode

For sports, Entertainment, Edtech, and E-shopping videos, mobile devices have a significant share. Video of 720p / 1080p with a 25 to 30 FPS frame rate provides a good experience. Being a small-screen 720p video with good quality can provide an excellent visual experience to users. The table below lists different classes of videos, Video bitrate with QBT & VMAF with on-mobile mode. The target mobile VMAF is 95.

The average bitrate for 720P is 947Kbps, where the range of bitrates is 493Kbps to 1525kbps. The average VMAF is 98.7, and min and max are 96.17 to 100. This mode of transcoding is suitable for Mobile and tablet devices. Here we can save bandwidth along with ensuring high video quality.

Economical Quality

For the Economical Mode, the target VMAF is 88 for 1080P Video. Most TV shows, movies, and News content can be streamed in this mode with good visual experience.

Target BR: 3000Kbps	Actual	1920×1080
	VMAF	BR in Kbps	BR Savings
TV Show1	92.04	828	72.40%
TV Show2	89.40	1224	59.20%
TV Show3	89.52	1613	46.23%
TV Show4	89.69	1912	36.27%
Movie1	87.92	1063	64.57%
Movie2	92.81	1960	34.67%
Movie3	92.45	2109	29.70%
Animation: Movie4	92.02	1757	41.43%
Animation: Movie5	90.21	1612	46.27%
Sports1	87.35	2875	4.17%
Sports2	92.78	2137	28.77%
Sports3	89.00	2071	30.97%
Sports4	94.59	2735	8.83%
Sports5	86.46	2071	30.97%
Average	90.45	1854.79	38.17%

For 1080P videos, the average bitrate is around 1860Kbps. The average VMAF is 90+. Min and max VMAFs are 86.46 and 92.81. Visually provides good quality.

Standard Quality

For the Standard Mode, the target VMAF is 92. In many cases, the transparent quality is considered as VMAF of 95, and +/- 3 VMAF points are nearly non-differential. Hence this mode provides very good quality for premium sports and shows.

Target BR: 4000Kbps	Actual	1920×1080
	VMAF	BR in Kbps	BR Savings
TV Show1	93.67	1037	74.08%
TV Show2	92.00	1588	60.30%
TV Show3	92.05	2082	47.95%
TV Show4	92.47	2605	34.88%
Movie1	91.83	1491	62.73%
Movie2	95.02	2474	38.15%
Movie3	94.68	2784	30.40%
Animation: Movie4	94.14	2261	43.48%
Animation: Movie5	92.64	2033	49.18%
Sports1	91.79	3314	17.15%
Sports2	97.10	3272	18.20%
Sports3	95.45	3183	20.43%
Sports4	96.90	3902	2.45%
Sports5	93.28	3158	21.05%
Average	93.79	2513.14	37.17%

For 1080P videos, the average bitrate is around 2514Kbps. The average VMAF is 93+. Min and Max VMAFs are 92.79 and 96.9. Visually provides very good quality.

Premium Quality

For the Premium Mode, the target VMAF is 95, the transparent video quality target. This mode provides original-quality video for premium movies, sports, and shows.

Target BR: 5000Kbps	5000 Kbps	1920×1080
	VMAF	BR in Kbps	BR Savings
TV Show1	95.84	1778	64.44%
TV Show2	94.96	2539	49.22%
TV Show3	94.76	3482	30.36%
TV Show4	95.08	4107	17.86%
Movie1	94.41	2131	57.38%
Movie2	97.49	3909	21.82%
Movie3	97.25	4093	18.14%
Animation: Movie4	96.75	3803	23.94%
Animation: Movie5	95.81	3286	34.28%
Sports1	94.40	4124	17.52%
Sports2	97.90	4072	18.56%
Sports3	96.86	3969	20.62%
Sports4	97.38	4452	10.96%
Sports5	95.44	3951	20.98%
Average	96.02	3549.71	29.01%

For 1080P videos, the average bitrate is around 3550Kbps. The average VMAF is 96+. Min and Max VMAFs are 94.76 and 97.49. Visually provides better quality.

HDR Quality

The desired quality video is very high for HDR video to retain the color information and brightness range of information. The target VMAF is 97. This mode provides the desired dynamic range of luminance and color gamut.

Target BR: 6000Kbps	Actual	1920×1080
	VMAF	BR in Kbps	BR Savings
TV Show1	97.51	5342	10.97%
TV Show2	96.96	5345	10.92%
TV Show3	96.89	5154	14.10%
TV Show4	96.87	5598	6.70%
Movie1	96.78	3818	36.37%
Movie2	98.08	5435	9.42%
Movie3	98.10	5557	7.38%
Animation: Movie4	97.78	5606	6.57%
Animation: Movie5	97.62	5503	8.28%
Sports1	97.52	5645	5.92%
Sports2	98.56	5567	7.22%
Sports3	97.92	5441	9.32%
Sports4	97.31	6207	-3.45%
Sports5	97.40	5429	9.52%
Average	97.52	5403.36	9.94%

For 1080P videos, the average bitrate is around 5400Kbps. The average VMAF is 97.5. Min and Max VMAFs are 96.78 and 98.56.

Summary and Conclusion

With QBT (Quality-Based Transcoding), the desired quality of the video can be achieved along with bitrate savings. For Mobile Video, a VMAF of 95+ can be achieved with an average of 950 Kbps bitrate for 720P.

For 1080P for TV devices Economical mode achieves an average VMAF 90+ of 1850 Kbps bitrate. The standard mode achieves a VMAF of 93+ with 2513 Kbps. The Premium mode provides an average VMAF of 96 with an average bitrate of 3550 Kbps.

The HDR mode provides an average VMAF of 97.52 with an average bitrate of 5400 Kbps.

We conclude that Quality-Based Transcoding has a simple interface, single pass optimized processing, target quality videos, and bandwidth savings, making it the best solution for VOD and Live streaming applications.