Video Optimization and Your Business
Cinova has the same goals and a similar pitch about its product and service line as does Beamr, but the approach and the customer base are very different. Cinova does not focus on a video-only solution, choosing instead to optimize both images and video. Additionally, it seems to have focused on retail customers.
Cinova Media’s CRUNCH solution is billed as being able to dramatically improve user experiences and encourage consumer engagement for ecommerce sites, video streaming services, and other media-rich applications by significantly reducing the delays, stutters, and lag associated with online video and photos. Ecommerce sites
deploying CRUNCH have seen dramatic increases in conversion of visitors to paying customers. According to Cinova, its perception optimized processing (POP) quantization reduces compressed files sizes up to 60% for photos and 40% for videos, while maintaining content fidelity and image quality.
Like other companies in the space, Cinova touts savings and improved user experience, though as mentioned earlier, its customer focus seems to be retailers and other large media entities looking to improve customer conversions through better load times of rich media.
An illustration of Cinova’s CRUNCH workflow
While offering up both video and image customer names, the technology seems to have grown out of the image optimization space, as Cinova’s numbers seem stronger there (highlighting more than 10 billion images that have been delivered to end users via CRUNCH). CRUNCH is fully cloud-based (AWS), and customers wishing to use it connect to it via REST APIs, with sample code and libraries available on Cinova’s website.
Company Name: Cinova
Target Customer: Ecommerce & CDNs
Proposed Savings: 40–60% on images, 40% on videos
Standout Feature: Works on images and videos
EuclidIQ is the most targeted of the bunch. Rather than creating a product for the media companies or those selling the content online, it has focused to the encoders themselves. Its offering, IQ264, is a licensable product (in SDK form) designed to augment existing H.264 encoders. IQ264 integrates directly into an encoder; it does not do preprocessing or postprocessing of existing files. It allows the encoder to continue to produce a compliant H.264 bitstream, which can be decoded by any standard H.264-compliant decoder. IQ264 is currently built on top of x264 and can be integrated with any H.264 encoder.
Perhaps the most interesting element from EuclidIQ was its attention to the objective and subjective testing of their end video results. Its testing methodology includes the following:
• Select a set of HD test videos, representative in terms of content (sports, nature, buildings, water) and video characteristics (complexity, motion).
• For each video, run both IQ264 and a baseline reference of x264.
• Evaluate encoder performance at multiple performance points below, at, and above video quality breakdown point.
• Run double-blind subjective tests to generate mean opinion scores (MOS) on a 1–5 scale for each video.
• Compute IQ264 gain vs. x264 via these mean opinion scores, measuring average bandwidth savings over common MOS interval.
A simplified example of EuclidIQ’s visual challenge was used to demonstrate IQ264 at Streaming Media East 2015.
IQ264 yields an average of more than 20% bandwidth savings as compared to the open source encoder x264 in both our controlled lab environment and even at the EuclidIQ Visual Challenge, which was in less-than-ideal lighting conditions. While lower than the claims made by Beamr and Cinova, EuclidIQ’s attention to testing gives some pretty solid background to reinforce those numbers.
Unlike the other solutions mentioned in this article, EuclidIQ's IQ264 is designed to work directly in the encoder's pipeline, not as a separate process.
The fact that it is integrated into primary encoders means one less step when readying files, which many will no doubt see as a benefit. EuclidIQ’s chief science officer Nigel Lee feels strongly that there is a misperception in the industry that video optimization using perceptual quality optimization (PQO) requires a pre- or postprocessing step. It has succeeded in avoiding additional processing steps and Lee feels that approach is a key differentiator and the reason it can provide lower bandwidths faster than other solutions.
EuclidIQ’s focus has been on eight key areas of the spatial and temporal image information—artifact detection, contrast sensitivity, differential motion, edge strength, global motion, luminance, structural similarity, and variance. EuclidIQ’s secret sauce with IQ264 is how it decides among the best of these in any given situation, offering the encoder a better way to encode any given frame or series of frames.
Company Name: EuclidIQ
Target Customer: Encoding Companies
Proposed Savings: 20% bandwidth savings with an additional 20% average processing times.
Standout Feature: Integrated directly in the encoder/doesnt require additional post- or pre-processing
Who Benefits from Video Optimization?
Despite the diversity in types of solutions presented here and the customers being catered to, there is one common aspect to who benefits in the optimization game: publishers and providers delivering very large amounts of content. In order for the cost and time it takes to do the work to pay off, it takes someone working at a large scale to get the results that make sense. This sentiment was fleshed out by Dan Rayburn in a blog post about Beamr last spring (go2sm.com/beamr). For those dealing with hundreds of files, the ROI could wind up being only a few dollars, but if the same scenario is applied to millions of files it may add up to tens of thousands of dollars in savings. So savings can surely add up, but to get to that point, you’ll be spending quite a bit on the process already.
It was no surprise to find that these companies’ solutions are radically different—my experience has been that many companies come to the same solution in very different ways—but what did surprise me was how different their targeted core customers were. While the teams may have customers across several industries, invariably they have a core publisher/provider market they focus on (the exception is EuclidIQ, which is focused on the encoder market). Because my background has been more on companies preparing media for OTT-type solutions, I expected to see a focus on that industry, and was most familiar with marketing language geared toward them, and so Cinova’s emphasis on ecommerce customers was eye-opening. For any company interested in an optimization solution, it is ultimately going to come down to testing it against your infrastructure and your content to see which is both the right fit and the right price.
Each of these companies has tackled the optimization problem in its own way and is targeting a widespread set of customers. That’s a good thing, because it is a diverse ecosystem. Few media companies have the same infrastructure or workflow, which means most aren’t looking for a cookie-cutter deployment. Their needs are as diverse as their customers and their offerings. The subjective nature of optimization will always leave wiggle room in what exactly gets touted, and that’s where each of these companies needs to make time to better inform the customers how and what they do if they expect to gain further traction. Optimization is also a constantly moving target. What we consider the best in optimization today will not hold up in 2 years, or possibly even sooner. So aggressively staying on top of improvements in file size and image quality is the challenge all companies in the space face if they want to stay relevant.
[This article appears in the September issue of Streaming Media magazine.]
Presenting a weather report on the industry's accelerated move toward cloud-based video acquisition and delivery.