Equirectangular Image Coding Distortion Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current coding techniques are inadequate for processing omnidirectional and multi-directional image content, leading to inefficient coding due to unaccounted image distortions, which result in failure to recognize redundancies in image content.
Innovation Solution
The proposed solution involves transforming reference picture data to counteract distortions caused by spherical projection formats, allowing for differential coding of input pixel blocks based on matches found in the transformed reference picture data, thereby aligning image content and reducing bandwidth consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If ordinary video coders process omnidirectional images without distortion compensation, then the coding process is simple, but coding efficiency deteriorates due to failure to recognize redundancies in distorted image content
Solution Approach 1:
The patent applies preliminary action by transforming reference picture data before coding to pre-compensate for spherical projection distortions. This allows the coder to work with corrected reference data that aligns with the current picture, enabling effective redundancy recognition without requiring complex distortion analysis during the coding process itself.
Solution Approach 2:
The patent introduces an intermediary transformation process that converts distorted reference picture data into a corrected form. This intermediary step acts as a mediator between the spherical projection format and the standard video coding process, allowing ordinary coders to handle omnidirectional content efficiently by working with transformed reference data.
2Productivity
If reference picture data is transformed to counteract spherical projection distortions, then coding efficiency improves by recognizing redundancies, but processing complexity increases
Solution Approach 1:
The patent applies local quality by performing distortion compensation locally for each reference picture block based on its specific displacement from the input pixel block. Rather than transforming the entire reference picture uniformly, the transformation parameters are adjusted locally to match the spherical projection geometry at each location, optimizing coding efficiency while managing processing complexity.
3Adaptability or versatility
If spherical projection formats are used for omnidirectional imaging, then field of view expands to capture 360 degrees, but image distortion increases making standard coding inadequate
Solution Approach 1:
The patent applies inversion by reversing the spherical projection transformation. Instead of projecting 3D scene points onto a 2D equirectangular grid (which causes distortion), the method transforms reference picture data in the opposite direction - from the distorted equirectangular coordinates back toward the spherical coordinate system - thereby compensating for distortions and restoring geometric accuracy for coding purposes.
Data Source
Figure 1~2
Figure 3(a)~3(c)
Figure 4
AI summary
Multi-directional image data often contains distortions of image content that cause problems when processed by video coders that are designed to process traditional, "flat" image content. Embodiments of the present disclosure provide techniques for coding multi-directional image data using such coders. For each pixel block in a frame to be coded, an encoder may transform reference picture data within a search window about a location of the input pixel block based on displacement respectively between the location of the input pixel block and portions of the reference picture within the search window. The encoder may perform a prediction search among the transformed reference picture data to identify a match between the input pixel block and a portion of the transformed reference picture and, when a match is identified, the encoder may code the input pixel block differentially with respect to the matching portion of the transformed reference picture. The transform may counter-act distortions imposed on image content of the reference picture data by the multi-directional format, which aligns the content with image content of the input picture. The techniques apply both for intra-coding and inter-coding.