Reading: Lin DCC’20 — CNN Based Fast Intra Mode Prediction for H.266/FVC Video Coding (Fast VVC)

To Reduce the Computational Complexity of Intra Coding

Outline

  1. Network Architecture
  2. Overall Approach
  3. Experimental Results

1. Network Architecture

Network Architecture
  • The CNN architecture comprises two convolutional layers and a fully connected layer.
  • The input including the neighbor pixels are fed into the network.
  • The output is a 67-dimensional vector for the 67 intra prediction modes.

2. Overall Approach

Overview flowchart of proposed method.
  • Only 16 × 16 blocks with the deep learning methodology were tested by JEM 7.0.
  • According to the CNN output, the top 5 modes are chosen to have full rate-distortion optimization (RDO) process.

3. Experimental Results

BD-PSNR, BDBR and Time Difference Compared to JEM-7.0 Without Fast Search
  • doFastSearch: The default SATD-based fast search.
  • Compared with the default fast search method doFastSearch in JEM 7.0, the proposed method can achieve averagely a 0.033% decrease in Bjøntegaard delta bit rate (BDBR) with only a slight increase in time.
  • Furthermore, the proposed method gains much improvement achieving a 0.097% decrease in BDBR over the default method when the tested videos are with moderate frame size such as classes B, C, D, E.

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store