Video Vectorization – reduce video size by about 95% for line-art based videos

栏目: IT技术 · 发布时间: 4年前

内容简介:Vectorly is developing a new kind of video compression technology, which uses computer vision and vector graphics to reduce bitrates for video content by an order of magnitude (or more) compared to HEVC, whileThis would be primarily effective for "vector f

Video Vectorization – reduce video size by about 95% for line-art based videos

Video Vectorization

Vectorly is developing a new kind of video compression technology, which uses computer vision and vector graphics to reduce bitrates for video content by an order of magnitude (or more) compared to HEVC, while improving video quality.

This would be primarily effective for "vector friendly" video content, which would include animations, screen-casts, many e-learning videos and potentially 3d gaming content.

By leveraging existing vector-graphics rendering capabilities on all devices, this codec wouldn't require end-users, OEMs or browsers to install special software to enable playback of these videos.

We are still in the early phases of developing this technology.

You can learn more about the technology in ourwhitepaper

The Core Idea

The core insight behind this project was that you could use vector-graphics based animations to simulate "videos" in a way that is indistinguishable from a traditional raster-graphics based video format such as an h264 video stream in an MP4 container.

Raster Graphics

Normal videos, like the ones you see on Netflix or YouTube, are just sequences of images which get updated quickly on the screen, to create the illusion of motion. Each image is composed of "pixels" - individual dots of color. Higher resolution means more pixels, better visual quality, and bigger file sizes.

Video Vectorization – reduce video size by about 95% for line-art based videos

Almost all video on the internet is of this format, known as "raster graphics". Video compression algorithms like h264 are just very efficient at using fewer data-points to reconstruct the pixels in any given frame, and at storing only the differences in pixels between frames of a video.

Vector Grapics video

In contrast, we use a concept called "vector-graphics" to render video. Instead of pixels, we represent everything on the screen using shapes, lines and curves, which can be represented as mathematical equations (vector graphics).

Video Vectorization – reduce video size by about 95% for line-art based videos

Using these mathematical equations, we can re-draw any arbitrary shape on the screen - from the letter "T" to Bart Simpson's head. Furthermore, by adding information such as color, position on the screen, and how they move or change shape over time, you can create whole videos - including entire episodes of the Simpsons, with just sequences of mathematical equations.

Why vectorization?

The core insight behind this project was that for a certain kind of "vector-friendly" video content, storing the video using vector graphics would be much more efficient than using raster graphics (in some cases, up to 2 orders of magnitude more efficient).

This idea is not substantively different from the idea of Flash based animations about 20 years ago. Why do this now?

No need for a decoder: Most devices now support SVG, HTML5, WebGL/OpenGL and/or some form of hardware-accelerated vector-graphics rendering. That lets you render vector-graphics content on any device without require end-users, OEMs or browsers to install special software to enable playback of vector-graphics content, and to achieve native-level performance by doing so. App developers would only need to include an appropriate library or SDK in their website or app to enable playback within native or 3rd player video players.

Computer vision: Our patented vectorization technology relies heavily on computer vision to convert raster-graphics videos to a vector format. Leveraging the advancement & commoditization of Computer Vision, and the ease of running batch computer-vision heavy tasks on the cloud, it's feasible to 'vectorize' large volumes of video at scale now, in a way that wasn't possible even 5 years ago.

Vector graphics video format

We are building a video-format based on existing standards (SVG, WebGL & OpenGL), extending it with Javascript to enable video features such as a timeline and key-frames. We package the resulting video data within an MP4 container, which can be streamed and distributed using existing video infrastructure (such as HLS/DASH, and DRM systems).

<video src="vectorized.mp4" type="video/svg">

We are pragmatic, and don't want to create a standard for the sake of creating a standard . To that end, we've created libraries and SDKs that enable playback of our vector-graphics videos using standard / native interfaces like so

<script src="vectorly.js">

<video src="vectorized.mp4" type="video/svg">
// This will work on all major browsers today

Demos / Proof of concept

Simpsons

Our first vectorized proof of concept for animations is a 17 second clip of the Simpsons locatedhere. Keep in mind, our technology is still at a very early stage, and this is much optimization work left to be done.

Khan Academy

Our technology also works very well for e-learning, and especially Khan Academy style content. You can find 30 second Khan Academy cliphere


以上就是本文的全部内容,希望本文的内容对大家的学习或者工作能带来一定的帮助,也希望大家多多支持 码农网

查看所有标签

猜你喜欢:

本站部分资源来源于网络,本站转载出于传递更多信息之目的,版权归原作者或者来源机构所有,如转载稿涉及版权问题,请联系我们

PHP经典实例(第3版)

PHP经典实例(第3版)

David Sklar、Adam Trachtenberg / 苏金国、丁小峰 / 中国电力出版社 / 2015-7 / 128.00

想要掌握PHP编程技术?或者想要学习如何完成一个特定的任务?那么一定要先看看《PHP经典实例(第3版)》。本书介绍了专门为PHP 5.4和5.5修订的350个经典技巧,并提供了丰富的示例代码。特别是对生成动态Web内容的解决方案做了全面更新,从使用基本数据类型到查询数据库,从调用RESTful API到测试和保护网站安全都有涵盖。 各个技巧都提供了示例代码,可以免费使用,另外还讨论了如何解决......一起来看看 《PHP经典实例(第3版)》 这本书的介绍吧!

HTML 压缩/解压工具
HTML 压缩/解压工具

在线压缩/解压 HTML 代码

RGB转16进制工具
RGB转16进制工具

RGB HEX 互转工具

XML 在线格式化
XML 在线格式化

在线 XML 格式化压缩工具