Back to Search Document details
18th Meeting: Teleconference, April 2020 2020-04-09 02:48
AHG16: On CCALF clipping
Abstract
In VVC CCALF process, an 8-tap filter is applied to luma sample to generate a residual correction for the chroma samples. At first, an offset value is generated from the luma samples and then the offset value is clipped. Then, the clipped offset value is added to the chroma sample to generate filtered output. Another clipping operation is performed to generate final filtered sample. This contribution proposes to remove the first clipping operation (i.e. clipping the offset value before sum) from the CCALF process. Following results are reported as compared to VTM-8.0.
JVET-R0128 AHG16: On CCALF clipping [M. G. Sarwer, Y. Ye, J. Luo (Alibaba)]

In VVC CCALF process, an 8-tap filter is applied to luma sample to generate a residual correction for the chroma samples. At first, an offset value is generated from the luma samples and then the offset value is clipped. Then, the clipped offset value is added to the chroma sample to generate filtered output. Another clipping operation is performed to generate final filtered sample. This contribution proposes to remove the first clipping operation (i.e. clipping the offset value before sum) from the CCALF process. Following results are reported as compared to VTM-8.0.

  • AI : 0.00% (Y), 0.00% (Cb), 0.00% (Cr)
  • RA : 0.00% (Y), 0.00% (Cb), 0.00% (Cr)
  • LB : 0.00% (Y), 0.00% (Cb), 0.00% (Cr)

In v2, the results of HDR sequences are added.

It is reported that the first clipping is never triggered in CTC.

One reason for introducing this clipping was saving memory for intermediate storage of luma data for later use in CCALF. There is also a conformance stream designed to check if the decoder implements the clipping.

No action was taken on this.

References:
JVET-Q2001
Decisions
No action was taken on this.
Citation