360-Degree Image Encoding Spherical Padding Motion Vector Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current encoding methods for 360-degree images are inefficient due to their reliance on techniques developed for ordinary images, lacking specificity for the unique characteristics of 360-degree images, which affects motion vector prediction and compression performance.
Innovation Solution
The proposed solution involves a padding operation that sets pixel values of the edge areas of 360-degree images by referencing adjacent areas on a sphere, allowing for accurate motion vector prediction and improved compression performance through the use of an encoding and decoding apparatus and method.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional encoding methods for ordinary images are used for 360-degree images, then the encoding process is simple, but the motion vector prediction accuracy is poor and compression performance is insufficient
Solution Approach 1:
The patent applies padding to edge areas of reference frames before motion estimation to pre-establish continuous pixel data across spherical boundaries. This preliminary action ensures that motion vectors for blocks near image edges can be accurately predicted by referencing pixels from adjacent areas on the sphere, thereby improving motion vector prediction accuracy without requiring complex post-processing
Solution Approach 2:
The patent treats the 360-degree image as a spherical projection where opposite edges are adjacent. By applying padding that wraps around the spherical boundary (e.g., left edge adjacent to right edge, top edge adjacent to bottom edge), the method extends the two-dimensional planar image into a three-dimensional spherical space, enabling accurate motion prediction across boundaries while maintaining computational efficiency
2Measurement precision
If padding is applied to edge areas by referencing adjacent areas on a sphere, then motion vector prediction accuracy improves, but the encoding complexity increases
Solution Approach 1:
The patent applies padding operations selectively only to edge areas of reference frames where spherical boundary effects occur, rather than processing the entire image uniformly. This localized approach improves motion vector prediction accuracy at critical boundary regions while minimizing additional computational complexity in the overall encoding process
Solution Approach 2:
The padding operation is performed as a preliminary step before motion estimation and compensation. By pre-establishing continuous pixel data at edges through spherical wrapping, the method enables standard motion estimation algorithms to work effectively on 360-degree images without requiring complex modifications to the core encoding pipeline
3Productivity
If ordinary image encoding methods are used, then the encoding process is straightforward, but compression performance is insufficient for 360-degree images
Solution Approach 1:
The patent applies spherical padding to reference frames before encoding to pre-resolve boundary discontinuities. This preliminary preparation enables more accurate motion vector prediction and reduces residual errors, thereby improving compression performance while adding only a single preprocessing step to the encoding pipeline
Solution Approach 2:
By conceptualizing the image as a spherical projection where opposite edges are adjacent, the patent enables motion vectors to reference pixels from wrapped-around areas. This dimensional transformation allows standard compression algorithms to achieve better performance on 360-degree images by exploiting the spherical topology without requiring fundamentally new encoding methods
Data Source
AI summary
An encoding apparatus comprises: a network interface for receiving an image; a memory for storing a command for encoding the image; and a processor for encoding the image according to the command, wherein the processor performs padding to set pixel values of a padding area corresponding to one edge area of the image by referring to the other edge area of the image adjacent to the one edge area on a sphere on which the image is projected, seeks a motion vector by referring to the pixel values of the padding area and encodes the image by referring to the motion vector, wherein the image is a 360-degree image.