360-Degree Panoramic Video Border Region Upsampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding techniques face challenges in efficiently encoding and decoding 360-degree panoramic video, particularly in handling border regions where sample values and variable values from opposite sides are not effectively utilized for prediction and entropy coding, leading to suboptimal image quality and increased bitrates.
Innovation Solution
The method involves reconstructing and encoding 360-degree panoramic source pictures by upsampling border regions using sample values and variable values from opposite sides, and determining reference regions that cross picture boundaries, enabling improved prediction and context-adaptive entropy coding for border regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional video coding techniques are used for 360-degree panoramic video, then the encoding and decoding process is simpler, but the image quality in border regions deteriorates and bitrate increases
Solution Approach 1:
The patent segments the panoramic picture into border regions and non-border regions, applying different processing techniques to each. Border regions utilize sample values and variable values from opposite sides through upsampling and context-adaptive entropy coding, while non-border regions use conventional processing. This segmentation allows targeted improvement of border region quality without unnecessarily complicating the entire encoding/decoding process.
Solution Approach 2:
The patent exploits the wraparound property of panoramic images by accessing sample values and variable values from the opposite side of the picture boundary. This dimensional approach treats the panoramic image as a continuous wraparound structure, allowing border regions to utilize information from across the entire panoramic field-of-view, thereby improving prediction accuracy and reducing bitrate in border regions without significantly increasing overall complexity.
2Measurement precision
If sample values from opposite sides are utilized for border region prediction, then prediction accuracy improves, but processing complexity increases
Solution Approach 1:
The patent applies enhanced processing techniques specifically to border regions where they are most beneficial, while conventional techniques are used for non-border regions. This local quality approach ensures that the increased processing complexity is concentrated only where it provides the greatest improvement in prediction accuracy, rather than applying complex processing uniformly across the entire picture.
Solution Approach 2:
The patent performs preliminary identification and segmentation of border regions before the actual encoding/decoding process. By pre-defining which regions will utilize opposite-side sample values and variable values, the system prepares the necessary data structures and processing paths in advance, reducing the real-time processing complexity during actual encoding and decoding operations.
3Quantity of substance
If context-adaptive entropy coding is applied to border regions using opposite side values, then bitrate is reduced, but encoding and decoding complexity increases
Solution Approach 1:
The patent changes the parameters used for entropy coding in border regions by incorporating opposite-side sample values and variable values into the context-adaptive entropy coding process. This parameter change allows the encoder to more accurately model the statistical properties of border region data, resulting in more efficient compression and reduced bitrate. The complexity increase is managed by applying these enhanced parameter changes only to border regions rather than the entire picture.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
There are disclosed various methods, apparatuses and computer program products for video encoding. In some embodiments the method comprises reconstructing a 360-degree panoramic source picture for inter-layer prediction; deriving an inter- layer reference picture from the 360-degree panoramic source picture, wherein the deriving comprises one or both of: upsampling at leasta part of the 360-degree panoramic source picture, wherein said upsampling comprises filtering samples of a border region of the 360-degree panoramic source picture using at least partly one or more sample values of an opposite side border region and/or one or more variable values associated with one or more blocks of the opposite side border region; determining a reference region that crosses a picture boundary of the 360-degree panoramic source picture, and including in the reference region one or both of the following: one or more sample values of an opposite side border region; one or more variable values associated with one or more blocks of the opposite side border region.