Intel and Badaboom* Video File Transcoding

Donwload Article

Download Intel and Badaboom Video File Transcoding [PDF 546KB]


Introduction

Intel® Quick Sync Video, built right into 2nd generation Intel® Core™ processors, is breakthrough hardware acceleration that lets the user complete in minutes what used to take hours. Create DVDs or Blu-ray discs, cover video files for your media plater, and convert video for upload to your favorite social networking sites - all in a flash.

Badaboom* uses Intel® Quick Sync Video technology to transcode video files in just minutes. Why do videos need to be transcoded? In order for a video to play back on a device such as a smartphone or a tablet, it needs to be formatted to correct specifications. With so many different devices out there, odds are low a video from a camcorder will automatically play on all of them. That's where Badaboom comes in: it transcodes video files to play on hundreds of the most popular devices available today-and it does so quickly and easily.

Courtesy: Elemental Technologies, Inc


Badaboom was built with four key cornerstones in mind, and these have been strengthened with its implementation of the Intel Media SDK:
  • Efficient use of system resources: Badaboom takes advantage of Intel® HD Graphics 2000 and 3000 series chips, offloading the CPU and allowing users to continue other tasks without unnecessarily slowing down their system.
  • Transcoding speed: Intel's Quick Sync Video technology enables Badaboom to transcode at several hundreds of frames per second.
  • Uncompromised video quality: Intel's Quick Sync H.264 encoder ensures Badaboom will continue to produce high quality outputs.
  • Ease of use and flexibility: Whether it is the default settings or advanced features, the Intel Media SDK supports the video processing parameters necessary for the flexible Badaboom interface to operate as users have come to expect.

Badaboom also utilizes video enhancement features, such as scaling, provided by Intel Quick Sync Video technology. More information about which video pre-processing filters are supported by the Media SDK can be found later in this paper.

The Intel® Media Software Development (MSDK) library exposes Intel Quick Sync Video technology to the application layer through three sets of functions: Decode, Video Pre Processing (VPP), and Encode. After Badaboom does any necessary demuxing of the input video file, the raw video frames are handed off to the Intel MediaSDK for the video processing. Badaboom performs video processing on a separate process from the demuxing, audio processing, and muxing, which allows it to integrate the Media SDK into its transcoding pipeline.

Badaboom uses asynchronous Decode, VPP and Encode functions to submit multiple frames to Intel MSDK to better and efficiently utilizes Intel® QSV.


Intel MSDK:
Mentioned earlier, the Intel Media SDK provides three set of functions to access Intel QSV features provided by 2nd generation of Intel® Core processors. These function are divided into following categories

  1. Decode
  2. VPP
  3. Encode

Intel MSDK Initialization:

The application initializes the MSDK session through MFXInit() function before it can use any MSDK functions. Each MSDK session can run exactly one instance of Decode, VPP and Encode functions. The application can start multiple sessions to process multiple frames in parallel. These multiple sessions can run independently or can be joined through to share the session data. The sessions can be joined through MFXJoinSession or MFXCloneSession.

mfxSession session, clone;
// to initialize a session

status = MFXInit(MFX_IMPL_AUTO, 0 , &session);

// to clone the session
status = MFXCloneSession(session, &clone);

//to close a session
Status = MFXClose(session);

MSDK Decode:

Media SDK Decode takes a compressed pure or elementary video stream as input and converts it to raw frames. The application can retrieve the decoded frames in display order or decoded order. The output frames are in NV12 color format. The input codecs supported by Intel® Media SDK are H.264, MPEG-2 and VC-1.

The Intel® Media SDK provides a very simple and easy to way to expose decode capabilities of Intel® Quick Sync Video technology. At high level, application needs to perform following steps:

  1. Initialize Decode Parameters: The application needs to initialize decode parameters before initializing decode processing. The application can call mfxVideoDecode_DecodeHeader() to obtain decode parameters This function parses the input bitstream and fills the decode parameters in mfxVideoParam structure.
  2. Get Surface Allocation Requirements: If two components are going to share surfaces, application needs to find out the surface allocation requirement. This can be done by calling MFXVideoDECODE_QueryIOSurf().
  3. Decoder Initialization: The application needs to initialize the decoder using MFXVideoDECODE_Init(). At this point, the Intel® Media SDK is ready to decode the frames.
  4. Frame Decoding: The application can decode the frames using MFXVideoDECODE_DecodeFrameAsync(). This is an asynchronous function and returns immediately without waiting for the result. Media SDK performs following four steps:
    • Checks the input data
    • Prepares a valid computation workload for hardware acceleration
    • Submits the workload for processing
    • Update the fields in the output structure as needed.
    The Media SDK returns a sync point after successful execution of asynchronous function. The application needs to use this sync point to synchronize the results using MFX_VideoCORE_SyncOperation().
  5. De-initialize the Decoder: After all frames decoded and retrieved de-initialize the decoder using MFXVideoDECODE_Close().

Video Pre-Processing

Intel® Quick Sync Video provides many Video Pre-Processing(VPP) filters:

  • Deinterlacing/Inverse Telecine
  • Cropping and Resizing
  • Color Conversion
  • Scene Edge Detection
  • Denoising
  • Detailing
  • Frame Rate Conversion
  • ProcAmp Filter
The application can access these filters using Intel® Media SDK functions. The VPP takes raw frames as input and produces raw frames as output. To utilize VPP filters of Intel® Quick Sync Video technology, the application needs to perform following steps:

  1. Get Surface Allocation Requirements: The application can retrieve the surface allocation requirement by calling MFXVideoVPP_QueryIOSurf() function.
  2. Initialize VPP: The application needs to initialize VPP before submitting frames for preprocessing. The initialization can be done using MFXVideoVPP_Init().
  3. Preprocess Frames: To preprocess a frame, the application uses MFXVideoVPP_RunFrameVPPAsync() function to submit a frame for preprocessing. If number of frames consumed at input is more than the output frames, VPP returns MFX_ERR_MORE_DATA for additional inputs. If number of frames at input is less than the output, VPP returns either MFX_ERR_MORE_SURFACE (if more outputs are ready) or MFX_ERR_NONE (if one output is ready).
  4. De-initialize the preprocessor: After video pre-processing is done, the applicant can de-initialize the preprocessor by calling MFXVideoVPP_Close().

Intel Media SDK Encode

MediaSDK Encode takes raw frames as input and compress them into a bitstream. Encode can process frames in Display order or in Encoded order. The encode supports variable bit rate (VBR) and constant Quantization Parameter (QP). The Encode output consists of one frame of a bitstream with the times stamp passed from the input frame. To encode frames using Intel Quick Sync Video technology, the application needs to perform following steps:

  1. Get Surface Allocation Requirements: If encoder is sharing surfaces with other components, then it needs to get the surface requirement. The application can retrieve this information using MFXVideoEncode_QueryIOSurf() function.
  2. Initialize Encoder: The application needs to initialize the encoder using MFXVideoEncode_Init() function.
  3. Encode Frames: To encode a frame, the application uses MFXVideoENCODE_EncodeFrameAsync() function. If application wants to set a new encoding configuration, it needs to retrieve any cached frames in encoder using MFXVideoENCODE_EncodeFrameAsync() function with a "NULL" input frame pointer. After that application needs to call MFXVideoENCODE_Reset() function with the new configuration.
  4. De-initialize the Encoder: After all frames are encoded and retrieved by the application, application can de-initialize the encoder using MFXVideoENCODE_Close() function.

About Elemental

Elemental Technologies is the leading provider of massively parallel video processing solutions for broadcast and online video customers. Elemental's products use off-the-shelf, programmable graphics processing units (GPUs) for compute-intensive video processing and conversion tasks. The product line is ideal for digital media workflows that require video encoding for Internet and mobile delivery, faster-than-real-time transcoding, or support for adaptive bit rate distribution. Elemental helps content owners and distributors deliver any video to any screen, at any time - all at once... it's Media Multiplied™. Founded in 2006, Elemental is headquartered in Portland, Oregon. To learn more about Elemental, visit www.elementaltechnologies.com.

For more complete information about compiler optimizations, see our Optimization Notice.