The Video Compression Problem

  • Compressed Video Delivery consumes some 80% of all Internet bandwidth
  • Individual companies annually spend hundreds of millions of dollars serving and storing compressed video
  • Each new video compression standard reduces bandwidth usage – but requires deployment of new decoders for that standard
  • What is needed is a means of reducing bandwidth usage of existing technology without changing either encoder or decoder

The Solution

ZPEG’s Pre-Processor increases video compressibility, thereby reducing the cost of storage and delivery while maintaining an equal visual experience.

The Pre-Processor improves the quality of compression when used in conjunction with existing technologies (MPEG-2, AVC, HEVC, VP-9), and requires no change to the encoder or decoder.

More about the Pre-Processor

ZPEG’s mainstream current product is a Pre-Processor that improves the quality of compression when used in conjunction with existing technologies (MPEG-2, AVC, HEVC, VP-9).

  • ZPEG pre-processing does not require any change to the decoder.
  • ZPEG pre-processing does a better job of removing temporal redundancies than is possible with standards-based compression.
  • ZPEG pre-processing is implemented using a human visual model that optimizes each specific asset (video), providing far lower bandwidths without objectionable artifacts (blocking, stair-stepping, ghosting).
Video compression

Adding Pre-Processor to Existing Workflows

Video compression
  • The ZPEG Pre-processor Seamlessly Drops into Existing Workflows
  • The ZPEG Pre-Processor works with MPEG-2, AVC, HEVC, VP-9

Seeing the difference in compression

Unprocessed

1. Side-by-side Overview

  • ALT_TEXT_OF_IMAGE

    4 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    6 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    Unprocessed & Uncompressed
    Download Video

  • Aggressive compression results in block artifacts, loss of detail, and loss of resolution and color fidelity. Click to big-ify.

2. Block Artifacts

  • ALT_TEXT_OF_IMAGE

    4 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    6 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    Unprocessed & Uncompressed
    Download Video

  • The compression artifacts become quite visible in the reflection of the building in the lake as the bandwidth is reduced.

3. Loss of Detail

  • ALT_TEXT_OF_IMAGE

    4 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    6 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    Unprocessed & Uncompressed
    Download Video

  • The fine detail in the wave crests are clearly visible in the uncompressed and 6 Mbs unprocessed video, but are lost in the 4 Mbs unprocessed video.

4. Loss of Resolution

  • ALT_TEXT_OF_IMAGE

    4 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    6 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    Unprocessed & Uncompressed
    Download Video

  • The shadow cast by the window canopy becomes steadily less distinguishable as bandwidth is reduced.

Pre-Processed

1. Side-by-side Overview

  • ALT_TEXT_OF_IMAGE

    4 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    6 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    4 Mbs pre-processed
    Download Video

  • ZPEG at 4 Mbs produces comparable quality to normal compression technology at 6 Mbs.

2. Less Blocking

  • ALT_TEXT_OF_IMAGE

    4 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    6 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    4 Mbs pre-processed
    Download Video

  • The compression artifacts in the reflection of the building in the lake are actually fewer in the pre-processed version at 3.8 Mbs than in the 6 Mbs unprocessed.

3. More Detail

  • ALT_TEXT_OF_IMAGE

    4 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    6 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    4 Mbs pre-processed
    Download Video

  • 4 Mbs The fine detail in the wave crests appears in the ZPEG 4 Mbs clip, but fails to appear in the unprocessed 4 Mbs sequence.

4. Enhanced Resolution

  • ALT_TEXT_OF_IMAGE

    4 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    6 Mbs unprocessed
    Download Video

  • ALT_TEXT_OF_IMAGE

    4 Mbs pre-processed
    Download Video

  • The shadow cast by the window canopy is quite clear in the pre-processed version at 4 Mbs, showing the additional detail made possible by ZPEG processing.

“It is not supported for smaller resolutions”

Specification of a planned viewing distance

gives the user control over the pre-processing compression boost

The equivalent viewing distance is specified in visiBels (vB), a logarithmic measure of perceptual distortion.

Perceptibility of stated compression advantages are reached by pre-processing 1080p HD content with a visiBel strength of -12 vB, well below the limit of visibility of such content at the viewing distance (-7 vB). Immersive 4K content is preprocessed at -18 vB.

How Does It Work?

  • ZPEG Pre-Compression implements the mathematical equivalent of the visual system physical processes (modulation threshold, nerve conduction speed, nerve system noise), which is in the case of MPEG compression systems implemented poorly or not at all.
  • The combination of these complementary technologies results in more effective compression

ZPEG pre-processing does a better job of removing temporal redundancies than is possible with standards-based compression. ZPEG is implemented using a human visual model that supports per-asset optimization, and far lower bandwidths can be obtained without objectionable artifacts (blocking, stair-stepping, ghosting).
The ZPEG Pre-processor is an application of the Human Visual Model to decorrelated transform space – the basic concept underlying ZPEG’s patent portfolio. The decorrelated transform domain supports theoretically ideal removal of imperceptible visual data. ZPEG’s Human Visual Model goes a step further by calibrating the model to a specified viewing distance, enabling the removal of all imperceptible data from the video stream at a desired visual quality.

Video compression
Video compression

A legacy compressor then processes a video stream that has been cleaned by removing imperceptible video content. The resulting content appears to be clearer than competing designs, resulting in a smaller representation of each frame.
While there is no theory governing the removal of error between frames, the ZPEG pre-processor is able to perform the removal in advance, following a well-defined theoretical model that guarantees optimality for a specified quality.