Banding prediction for video encoding

Patent metadata

Field	Value
Publication number	US-10034005-B2
Application number	US-201514732551-A
Country	US
Kind code	B2
Filing date	Jun 5, 2015
Priority date	Jun 5, 2015
Publication date	Jul 24, 2018
Grant date	Jul 24, 2018

Abstract

A system and method for facilitating inhibiting banding in video data, in part by anticipating when banding may occur given certain encoder parameters. An example method includes receiving an input stream of video data; extracting feature information characterizing the stream of video data; using a feed-forward neural network to process the feature information to estimate when a particular block of the stream of video data will exhibit an artifact when encoded using certain Quantization Parameters (QPs); incorporating the indication into metadata associated with the stream of video data; and transferring the stream of video data and metadata to a video encoder. In a more specific embodiment, the feature information includes color information and texture information, and the neural network includes a feed-forward neural network that includes a classifier with a sigmoid activation function, and which has been trained using a cross-entropy cost function.

First claim

We claim:

1. A method for facilitating inhibiting banding in video data, the method comprising:
receiving an input stream of video data;
using a feature extractor module to extract, from the stream of video data, feature information including color information and texture information characterizing one or more features of a block of the video data;
using a neural network to process the feature information to provide an indication of likelihood of whether the block of the stream of video data will exhibit an artifact including a banding artifact when the block is subsequently encoded using a video encoder;
incorporating the indication into metadata associated with the stream of video data; and
transferring the stream of video data and metadata to the video encoder;
wherein a banding artifact prediction is associated with one or more Quantization Parameters (QPs) associated with the block; and
wherein using a neural network further includes employing the neural network to predict whether a banding artifact will occur when the stream of video data is processed by the video encoder when the one or more QPs are used for the block.

2. The method of claim 1 , wherein the neural network includes a feed-forward neural network.

3. The method of claim 2 , wherein the feed-forward neural network includes a classifier with a sigmoid activation function.

4. The method of claim 1 , wherein incorporating further includes adjusting one or more encoder parameters and/or applying pre-processing so as to eliminate or reduce the banding artifact for blocks where the neural network predicts the banding artifact to occur otherwise.

Assignees

• Xu Xun
• Ye Peng
• Sony Corp

Inventors

• Xu Xun
• Ye Peng

Classifications

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N3/048

  Activation functions · CPC title

• H04N19/117

  Filters, e.g. for pre-processing or post-processing (sub-band filter banks H04N19/635) · CPC title

• H04N19/176 Primary

  the region being a block, e.g. a macroblock · CPC title

• H04N19/103

  Selection of coding mode or of prediction mode · CPC title