Browse Prior Art Database

A Highly Efficient Parallel Intra-Prediction Method

IP.com Disclosure Number: IPCOM000242559D
Publication Date: 2015-Jul-24
Document File: 4 page(s) / 68K

Publishing Venue

The IP.com Prior Art Database

Abstract

For most popular video standards, such as H.264/H.265, intra-predictions of blocks are difficult to process in parallel due to the dependency of neighboring blocks. In this document, one highly efficient parallel intra-prediction method is proposed based on partition and reorder of block pixels.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 54% of the total text.

A Highly Efficient Parallel Intra-Prediction Method

Abstract

For most popular video standards, such as H.264/H.265, intra-predictions of blocks are difficult to process in parallel due to the dependency of neighboring blocks.  In this document, one highly efficient parallel intra-prediction method is proposed based on partition and reorder of block pixels.

What’s the problem?

In some popular video standards, there are strong dependencies between internal blocks for intra-prediction process. The processing of current block must refer to pixels in neighboring blocks (includes top and left boundaries).  As a result, intra-prediction of block must be processed one by one based on strict scan order. 

What’s the solution?

This method implements one parallel intra-prediction through partition and reorder of pixels in the processed block.

The strategies include:

·         Partition all pixels in one processed block with two set of pixels:

The first set contain referenced pixels which will be referenced by neighboring blocks for intra prediction;

The second set contain non-referenced pixels which won’t be referenced by neighboring blocks

One example of pixels partition for H.265 is:

Pixels in right and bottom boundary in processed block are classified into the first set, and other pixels belong to second set.

·         The intra prediction and reconstruction of the first set of pixels will be done before the second set of pixels in the same block

·         The intra prediction and reconstruction of the first set of pixels in current block only depend on the first set of pixels in neighboring blocks in terms of dependency defined by corresponding video encode standard

·         The intra prediction and reconstruction of the second set of pixels in one block are processed without any blocking or delay due to above partition

·         The first set of pixels can be split into more sub-sets in terms of dependency relation and the process order of sub-set can be changed accordingly

·         The dependency can be minimized through checking the current reference direction

·         The unit size processed by intra processor can be configured with more flexible size based on above partition and reorder

Work flow for single intra prediction processor:

Some feasible extension of the method

a)      For the first set of pixels, more sub-set can...