# US 12,604,005 B2 · Video decoding method, video encoding method, and video decoder

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,604,005 B2
- Document ID: US12604005B2
- Status: Granted
- Kind: B2
- Application number: 18/950,067
- Filed: 2024-11-16
- Published: 2025-03-06
- Granted: 2026-04-14
- Earliest claimed: 2022-05-18
- Assignees: Guangdong Oppo Mobile Telecommunications Corp., Ltd.
- Applicants: Guangdong Oppo Mobile Telecommunications Corp., Ltd.
- Inventors: Zhenyu Dai
- Web page: https://patentgrep.com/patent/US12604005B2

## CPC

- H04N 19/126: Details of normalisation or weighting functions, e.g. normalisation matrices or variable uniform quantisers
- H04N 19/117: Filters, e.g. for pre-processing or post-processing (sub-band filter banks H04N19/635)
- H04N 19/124: Quantisation
- H04N 19/154: Measured or subjectively estimated visual quality after decoding, e.g. measurement of distortion (use of rate-distortion criteria H04N19/147)
- H04N 19/176: the region being a block, e.g. a macroblock
- H04N 19/82: involving filtering within a prediction loop
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION

## Abstract

Embodiments of the present disclosure provide video decoding method, a video encoding method, and a video decoder. A decoding end operates as follows. A bitstream is decoded and a reconstructed block of a current block is determined. The bitstream is decoded and a first quantization parameter (QP) of the current block is determined, where the first QP is determined based on a second QP predicted through a QP prediction model. The reconstructed block of the current block is filtered with a neural network-based filter according to the first QP, to obtain a filtered reconstructed block.

## Claims

### Claim 1

A video decoding method, comprising:
decoding a bitstream and determining a reconstructed block of a current block;
decoding the bitstream and determining a first quantization parameter (QP) of the current block, wherein the first QP is determined based on a second QP predicted through a QP prediction model; and
filtering, according to the first QP, the reconstructed block of the current block with a neural network-based filter, to obtain a filtered reconstructed block.

### Claim 2 (depends on 1)

The method of claim 1, wherein determining the first QP of the current block comprises:
decoding the bitstream to obtain the first QP of the current block.

### Claim 3 (depends on 1)

The method of claim 1, wherein determining the first QP of the current block comprises:
predicting the second QP of the current block through the QP prediction model; and
determining the first QP according to the second QP.

### Claim 4 (depends on 3)

The method of claim 3, wherein determining the first QP according to the second QP comprises:
determining the first QP according to the second QP and a preset base quantization parameter (BQP).

### Claim 5 (depends on 4)

The method of claim 4, wherein determining the first QP according to the second QP and the preset BQP comprises:
determining a distortion corresponding to the second QP;
determining a distortion corresponding to the preset BQP; and
determining, according to the distortion corresponding to the second QP and the distortion corresponding to the preset BQP, the first QP from the second QP and the preset BQP.

### Claim 6 (depends on 5)

The method of claim 5, wherein determining, according to the distortion corresponding to the second QP and the distortion corresponding to the preset BQP, the first QP from the second QP and the preset BQP comprises:
determining a QP with a minimum distortion from the second QP and the preset BQP as the first QP.

### Claim 7 (depends on 5)

The method of claim 5, wherein determining the distortion corresponding to the second QP comprises:
filtering, according to the second QP, the reconstructed block of the current block with the neural network-based filter, to obtain a first filtered reconstructed block; and
determining the distortion corresponding to the second QP according to the first filtered reconstructed block and a reconstructed value of the current block.

### Claim 8 (depends on 7)

The method of claim 7, wherein filtering, according to the second QP, the reconstructed block of the current block with the neural network-based filter, to obtain the first filtered reconstructed block comprises:
inputting concatenation of the reconstructed value of the current block and the second QP into the neural network-based filter for filtering, to obtain the first filtered reconstructed block.

### Claim 9 (depends on 5)

The method of claim 5, wherein determining the distortion corresponding to the preset BQP comprises:
filtering, according to the preset BQP, the reconstructed block of the current block with the neural network-based filter, to obtain a second filtered reconstructed block; and
determining the distortion corresponding to the preset BQP according to the second filtered reconstructed block and a reconstructed value of the current block.

### Claim 10 (depends on 9)

The method of claim 9, wherein filtering, according to the preset BQP, the reconstructed block of the current block with the neural network-based filter, to obtain the second filtered reconstructed block comprises:
inputting concatenation of the reconstructed value of the current block and the preset BQP into the neural network-based filter for filtering, to obtain the second filtered reconstructed block.

### Claim 11 (depends on 1)

The method of claim 1, wherein determining the first QP of the current block comprises:
decoding the bitstream to obtain a first flag, wherein the first flag indicates whether the first QP of the current block is the second QP predicted through the QP prediction model;
in response to the first flag indicating that the first QP of the current block is the second QP predicted through the QP prediction model, predicting the second QP of the current block through the QP prediction model; and
determining the second QP as the first QP.

### Claim 12 (depends on 11)

The method of claim 11, further comprising:
in response to the first flag indicating that the first QP of the current block is not the second QP predicted through the QP prediction model, determining a preset BQP as the first QP of the current block.

### Claim 13 (depends on 11)

The method of claim 11, wherein decoding the bitstream to obtain the first flag comprises:
decoding the bitstream to obtain a second flag, wherein the second flag indicates whether a first QP of at least one block in a current picture is the second QP predicted through the QP prediction model; and
in response to the second flag indicating that the first QP of the at least one block in the current picture is the second QP predicted through the QP prediction model, decoding the bitstream to obtain the first flag.

### Claim 14 (depends on 13)

The method of claim 13, further comprising:
in response to the second flag indicating that a first QP of the current picture is not the second QP predicted through the QP prediction model, determining a preset BQP as the first QP of the current block.

### Claim 15 (depends on 13)

The method of claim 13, wherein decoding the bitstream to obtain the second flag comprises:
decoding the bitstream to obtain a third flag, wherein the third flag indicates whether the QP prediction model is allowed to be used for a current sequence for QP prediction; and
in response to the third flag indicating that the QP prediction model is allowed to be used for the current sequence for QP prediction, decoding the bitstream to obtain the second flag.

### Claim 16 (depends on 3)

The method of claim 3, wherein predicting the second QP of the current block through the QP prediction model comprises:
inputting the reconstructed block of the current block and at least one preset third QP into the QP prediction model for QP prediction, to obtain the second QP.

### Claim 17 (depends on 16)

The method of claim 16, wherein inputting the reconstructed block of the current block and the at least one preset third QP into the QP prediction model for QP prediction, to obtain the second QP comprises:
inputting concatenation of the reconstructed block of the current block and the at least one preset third QP into the QP prediction model for QP prediction, to obtain the second QP.

### Claim 18 (depends on 16)

The method of claim 16, wherein the at least one preset third QP comprises at least one of: a BQP, a slice quantization parameter (SQP), or a QP of a reference block of the current block.

### Claim 19

A video encoding method, comprising:
reconstructing a current block to obtain a reconstructed block of the current block;
predicting a second quantization parameter (QP) of the current block through a QP prediction model;
determining a first QP of the current block according to the second QP; and
filtering, according to the first QP, the reconstructed block of the current block with a neural network-based filter, to obtain a filtered reconstructed block.

### Claim 20

A video decoder, comprising:
at least one processor; and
a memory coupled to the at least one processor and storing at least one computer executable instruction thereon which, when executed by the at least one processor, causes the at least one processor to:
decode a bitstream and determine a reconstructed block of a current block;
decode the bitstream and determine a first quantization parameter (QP) of the current block, wherein the first QP is determined based on a second QP predicted through a QP prediction model; and
filter, according to the first QP, the reconstructed block of the current block with a neural network-based filter, to obtain a filtered reconstructed block.