Check patentability & draft patents in minutes with Patsnap Eureka AI!

Method and device for two-way template motion vector fine-tuning for video encoding and decoding

A video encoding and decoding, motion vector technology, applied in the field of fine-tuning the motion of bidirectional prediction blocks

Active Publication Date: 2022-03-08
HFI INNOVATION INC
View PDF2 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The fine-tuning search range is limited to within ±1 pixel

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Method and device for two-way template motion vector fine-tuning for video encoding and decoding
  • Method and device for two-way template motion vector fine-tuning for video encoding and decoding
  • Method and device for two-way template motion vector fine-tuning for video encoding and decoding

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0081] The main concept of Embodiment 1 is that if the motion vector difference after fine-tuning is less than a threshold, the iteration will be terminated. An example of a specific algorithm can be as Figure 9 As shown, Target_context_value=L0_MV_difference+L1_MV_difference. For example, L0_MV_difference may be the difference between the original L0 MV and the fine-tuned MV. Likewise, the L1MV difference can be calculated. According to the following equation, L0_MV_difference and L1_MV_difference can be calculated:

[0082] L0_MV_difference = abs("original L0 MV x part" - "fine tuned L0 MV x part") + abs("original L0 MV y part" - "fine tuned L0 MV y part")

[0083] L1_MV_difference = abs("original L1 MV x part" - "fine tuned L1 MV x part") + abs("original L1 MV y part" - "fine tuned L1 MV y part")

[0084] In the above equation, abs() is the absolute value. The "x part" is the horizontal part of the MV. "y-section" is the vertical section of the MV.

Embodiment 2

[0086] The main concept of embodiment 2 is that if the pixel difference between the current reference block and the fine-tuning reference block is smaller than some threshold, the iteration will be ended. The specific algorithm can be as Figure 9 As shown, Target_context_value=L0_block_difference+L1_block_difference. For example, L0_block_difference can be calculated as the difference between the original L0 reference block and the L0 fine-tuning block. Similarly, L1_block_difference can be calculated. According to the following procedure, L0_block_difference and L1_block_difference can be calculated:

[0087] L0_block_difference = SAD between L0 reference block and L0 trimming block

[0088] L1_block_difference = SAD between L1 reference block and L1 fine-tuning block

[0089] In the above equation, SAD corresponds to the sum of absolute differences, and the L0 reference block and the L0 fine-tuning block are defined as Figure 9 shown.

Embodiment 3

[0091] The main concept behind Example 3 is that if some combination of motion vector difference and pixel difference is less than a threshold, the iteration will be ended. An example of a specific algorithm for implementing this embodiment may be as Figure 9 As shown, Target_context_value=a*(L0_MV_difference+L1_MV_difference)+b*(L0_block_difference+L1_block_difference), where a and b are weighting factors.

[0092] Sub-PU-based two-way template MV fine-tuning

[0093] In order to improve encoding and decoding efficiency, the present invention discloses sub-PU-based two-way template MV fine-tuning. The main idea behind this method is to split the PU into sub-PUs and then perform bi-directional template MV fine-tuning in the sub-PU layer. An example of the specific operation of this method can be as Figure 10 , where only operations related to the L0 portion are shown. Similarly, operations related to the L1 part can be derived. Figure 10 The exemplary procedure in incl...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a fine-tuning method and device using bidirectional template motion vectors to improve codec efficiency or reduce complexity. According to one approach, bidirectional template motion vector fine-tuning is applied to the current block if the size associated with the current block is greater than a threshold. Otherwise, no bi-directional template motion vector fine-tuning is applied to the current block. In another approach, bi-directional template MV fine-tuning is implicitly turned on or off based on two reference blocks. According to yet another approach, bi-directional template motion vector fine-tuning will be performed on the sub-block level. According to yet another method, bi-directional template motion vector fine-tuning is performed based on the motion vectors selected from the candidates of the high-level motion vector prediction candidate list. According to yet another method, bi-directional template motion vector fine-tuning uses a modified template to fine-tune motion vectors.

Description

[0001] 【cross reference】 [0002] This application claims priority from U.S. Provisional Patent Application No. 62 / 439,203, filed December 27, 2016, U.S. Provisional Patent Application No. 62 / 439,207, filed December 27, 2016 , U.S. Provisional Patent Application No. 62 / 445,831 filed on January 13, 2017, U.S. Provisional Patent Application No. 62 / 455,625 filed on Feb. 7, 2017, U.S. Provisional Patent Application No. 62 / 455,625 filed on Feb. 21, 2017 The proposed U.S. Provisional Patent Application No. 62 / 461,300 and the U.S. Provisional Patent Application No. 62 / 469,048 filed on March 09, 2017. The aforementioned US Provisional Patent Application is hereby incorporated by reference in its entirety. 【Technical field】 [0003] The present invention relates to motion compensation using Bilateral Template MV Refinement (BTMVR) to fine tune the motion of a bidirectionally predicted block. Specifically, the present invention relates to different techniques related to the two-way te...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): H04N19/51
CPCH04N19/105H04N19/176H04N19/136H04N19/577H04N19/533H04N19/139H04N19/573
Inventor 陈俊嘉徐志玮庄子德陈庆晔黄毓文
Owner HFI INNOVATION INC
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com