IP Library › Granted Patent US 12,506,883
Granted Patent B2
US 12,506,883 · App. 18/207,587 · Granted Dec 23, 2025

Template matching for multiple reference line intra prediction

Inventors: Xin Zhao (Santa Clara, CA); Guichun Li (San Jose, CA); Lien-Fei Chen (Hsinchu, TW); Shan Liu (San Jose, CA)
Assignee: Tencent America LLC
H04N19/184H04N19/176H04N19/593
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,506,883
App. No.
18/207,587
Granted
Dec 23, 2025
Kind
B2
Abstract

Processing circuitry receives a video bitstream comprising a current coding block in a picture coded in a template-based multiple reference line intra prediction (TMRL) mode. The processing circuitry determines combinations of reference lines and intra prediction modes for the current coding block. A first combination in the combinations includes a first reference line and a first intra prediction mode. For the first combination, the processing circuitry calculates template entry delta values respectively associated with entries of a template of the current coding block based on the first reference line and the first intra prediction mode, and calculates a first template matching cost value by unequal weighting of the template entry delta values. The processing circuitry sorts the combinations according to template matching cost values respectively associated with the combinations, selects a specific combination from the sorted combinations, and reconstructs the current coding block according to the specific combination.

Claims (49)

1 . A method of video decoding, comprising:

receiving a video bitstream comprising a current coding block in a picture;

determining that the current coding block is coded in a template-based multiple reference line intra prediction (TMRL) mode;

determining combinations of reference lines and intra prediction modes for an intra prediction of the current coding block, a first combination in the combinations comprising a first reference line in the reference lines and a first intra prediction mode in the intra prediction modes, the first reference line being a non-adjacent reference line;

for the first combination, calculating template entry delta values respectively associated with entries of a template of the current coding block, the template of the current coding block being in an adjacent reference line, each of the template entry delta values being equal to a difference value between (i) an entry at a respective position in a reference template that is determined by the first reference line and the first intra prediction mode and (ii) an entry at the respective position in the template of the current coding block;

calculating a first template matching cost value associated with the first combination based on weighted template entry delta values by unequally weighting of the template entry delta values;

ordering the combinations of the reference lines and the intra prediction modes according to template matching cost values respectively associated with the combinations, the template matching cost values associated with the combinations including the first template matching cost value associated with the first combination;

selecting a specific combination from the ordered combinations; and

reconstructing the current coding block according to the specific combination.

2 . The method of claim 1 , wherein the template of the current coding block comprises neighboring samples to the current coding block as the entries, and the method comprises:

determining the reference template in the first reference line corresponding to the template of the current coding block based on the first intra prediction mode, the reference template comprising reference entries respectively corresponding to the entries of the template.

3 . The method of claim 2 , further comprising:

calculating a first template entry delta value as a difference between a first entry in the entries of the template and a first reference entry in the reference entries of the reference template, the first reference entry corresponding to the first entry; and

determining a weight for weighting the first template entry delta value based on at least the first entry and the first reference entry.

4 . The method of claim 3 , further comprising:

determining the weight for weighting the first template entry delta value based on the first entry, the first reference entry and one or more adjacent reference entries of the first reference entry.

5 . The method of claim 3 , further comprising:

determining the weight for weighting the first template entry delta value based on the first template entry delta value and one or more template entry delta values associated with adjacent entries of the template.

6 . The method of claim 3 , further comprising:

determining the weight for weighting the first template entry delta value based on at least a first gradient value at the first entry and/or a second gradient value at the first reference entry.

7 . The method of claim 2 , further comprising:

calculating the first template matching cost value associated with the first combination based on a portion of the template.

8 . The method of claim 7 , further comprising:

calculating the first template matching cost value based on an above template that includes above neighboring samples of the current coding block.

9 . The method of claim 7 , further comprising:

calculating the first template matching cost value based on a left template that includes left neighboring samples of the current coding block.

10 . A method of video encoding, comprising:

determining that a current coding block in a current picture is to be coded in a template-based multiple reference line intra prediction (TMRL) mode;

determining combinations of reference lines and intra prediction modes for an intra prediction of the current coding block, a first combination in the combinations comprising a first reference line in the reference lines and a first intra prediction mode in the intra prediction modes, the first reference line being a non-adjacent reference line;

for the first combination, calculating template entry delta values respectively associated with entries of a template of the current coding block, the template of the current coding block being in an adjacent reference line, each of the template entry delta values being equal to a difference value between (i) an entry at a respective position in a reference template that is determined by the first reference line and the first intra prediction mode and (ii) an entry at the respective position in the template of the current coding block;

calculating a first template matching cost value associated with the first combination based on weighted template entry delta values by unequally weighting of the template entry delta values;

ordering the combinations of the reference lines and the intra prediction modes according to template matching cost values respectively associated with the combinations, the template matching cost values associated with the combinations including the first template matching cost value associated with the first combination;

selecting a specific combination from the ordered combinations; and

encoding the current coding block according to the specific combination.

11 . The method of claim 10 , wherein the template of the current coding block comprises neighboring samples to the current coding block as the entries, and the method comprises:

determining the reference template in the first reference line corresponding to the template of the current coding block based on the first intra prediction mode, the reference template comprising reference entries respectively corresponding to the entries of the template.

12 . The method of claim 11 , further comprising:

calculating a first template entry delta value as a difference between a first entry in the entries of the template and a first reference entry in the reference entries of the reference template, the first reference entry corresponding to the first entry; and

determining a weight for weighting the first template entry delta value based on at least the first entry and the first reference entry.

13 . A method of processing visual media data, the method comprising:

processing a bitstream of the visual media data according to a format rule, wherein:

the bitstream includes coded information of a current coding block in a picture, the current coding block being coded in a template-based multiple reference line intra prediction (TMRL) mode; and

the format rule specifies that:

combinations of reference lines and intra prediction modes are determined for an intra prediction of the current coding block, a first combination in the combinations comprising a first reference line in the reference lines and a first intra prediction mode in the intra prediction modes, the first reference line being a non-adjacent reference line;

for the first combination, template entry delta values are calculated respectively associated with entries of a template of the current coding block, the template of the current coding block being in an adjacent reference line, each of the template entry delta values being equal to a difference value between (i) an entry at a respective position in a reference template that is determined by the first reference line and the first intra prediction mode and (ii) an entry at the respective position in the template of the current coding block;

a first template matching cost value associated with the first combination is calculated based on weighted template entry delta values by unequally weighting of the template entry delta values;

the combinations of the reference lines and the intra prediction modes are ordered according to template matching cost values respectively associated with the combinations, the template matching cost values associated with the combinations including the first template matching cost value associated with the first combination;

a specific combination is selected from the ordered combinations; and

the current coding block is processed according to the specific combination.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 8, 2023
From: ZHAO, XIN; LI, GUICHUN; CHEN, LIEN-FEI; LIU, SHAN
To: TENCENT AMERICA LLC
Reel/Frame 063900/0270 →
Continuity (2)
Provisional Application 63388909 · Jul 13, 2022
Related Publication 20240022738A1 · Jan 18, 2024
References Cited (15)
US 9363511B2 · Zhang et al. · 2016 [cited by applicant]
US 9769472B2 · Liu et al. · 2017 [cited by applicant]
US 10419754B1 · Zhao · 2019 [cited by examiner]
US 20230217019A1 · Teng · 2023 [cited by examiner]
JP 2012532501A · 2012 [cited by applicant]
JP 2019535211A · 2019 [cited by applicant]
JP 2021518059A · 2021 [cited by applicant]
High Efficiency Video Coding, Rec. ITU-T H.265 v4 Dec. 2016, pp. 1-664. [cited by applicant]
B. Bross, J. Chen, S. Liu, “Versatile Video Coding (Draft 2)”, ISO/IEC JTC1/SC29/WG11 JVET-K1001, Jul. 2018, pp. 1-139. [cited by applicant]
Zhang, Ximin, et al. Non-CE6: Intra mode coding with fixed length binarization. Joint Collaborative Team on Video Coding (JCT-VC) of ITU-T SG16 WP3 and ISO/IEC JTC1/SC29/WG11, 8th Meeting, JCTVC-H0435. Feb. 2012, pp. 1-… [cited by applicant]
Y.-J. Chang et al., “Multiple Reference Line Coding for Most Probable Modes in Intra Prediction,” 2019 Data Compression Conference (DCC), 2019, pp. 559-559, doi: 10.1109/DCC.2019.00071. [cited by applicant]
B. Bross et al., CE3: Multiple reference line intra prediction (Test 1.1.1, 1.1.2, 1.1.3 and 1.1.4), document JVET-L0283 of JVET, Oct. 2018, pp. 1-7. [cited by applicant]
L. Xu, Y. Yu, H. Yu, D. Wang, “Non-EE2: Template-based multiple reference line intra prediction,” Joint Video Experts Team (JVET) of ITU-T SG 16 WP 3 and ISO/IEC JTC 1/SC 29, Doc. JVET-AA0120, 27th Meeting, by teleconfe… [cited by applicant]
Office Action received for Japanese Patent Application No. 2024-547009, mailed on Jul. 15, 2025, 23 pages (13 pages of English Translation and 10 pages of Original Document). [cited by applicant]
Xu et al., “Non-EE2: Template-based multiple reference line intra prediction”, OPPO, Joint Video Experts Team (JVET) of ITU-T SG 16 WP 3 and ISO/IEC JTC 1/SC 29 27th Meeting, by teleconference, Jul. 13-22, 2022, Documen… [cited by applicant]