Edge-Aware Geometry Quantization for Multiview Rendering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multiview images suffer from perceptible visual artifacts due to quantization errors when rendered at intermediate viewpoints, particularly when higher frequency components are quantized using a higher step, leading to issues like 'flying points' and object bloating.
Innovation Solution
Generate an edge feature atlas to indicate edge or discontinuity samples, using a smaller quantization step for samples not at edges, thereby reducing quantization errors and visual artifacts in reconstructed scenes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If a higher quantization step is used to quantize higher frequency components of a geometry atlas, then compression efficiency is improved, but visual artifacts become perceptible in reconstructed scenes at intermediate viewpoints
Solution Approach 1:
The patent applies local quality by differentiating quantization treatment based on spatial location. Edge samples identified in the edge feature map receive a first (smaller) quantization step, while non-edge samples receive a second (larger) quantization step. This localized differentiation preserves edge sharpness and reduces visual artifacts in critical regions while maintaining compression efficiency in non-critical regions.
Solution Approach 2:
The patent changes the quantization parameter dynamically based on edge detection results. The quantization step size is adjusted according to the edge feature map, transitioning from a uniform quantization approach to a variable quantization approach where the parameter (quantization step) changes spatially to optimize both compression and visual quality.
2Device complexity
If uniform quantization is applied to all samples in the geometry atlas, then processing complexity is reduced, but edge sharpness and visual quality deteriorate
Solution Approach 1:
The patent performs preliminary action by generating an edge feature map before the quantization process. This edge detection step identifies which samples require special treatment, allowing the subsequent quantization process to apply appropriate differentiation. The preliminary edge analysis enables targeted processing that improves edge sharpness without requiring complex real-time adjustments during quantization.
3Object-affected harmful factors
If quantization errors are reduced by using smaller quantization steps, then visual quality is improved, but data compression efficiency decreases
Solution Approach 1:
The patent applies local quality by differentiating quantization treatment based on spatial location. Edge samples identified in the edge feature map receive a first (smaller) quantization step, while non-edge samples receive a second (larger) quantization step. This localized differentiation preserves edge sharpness and reduces visual artifacts in critical regions while maintaining compression efficiency in non-critical regions.
Solution Approach 2:
The patent changes the quantization parameter dynamically based on edge detection results. The quantization step size is adjusted according to the edge feature map, transitioning from a uniform quantization approach to a variable quantization approach where the parameter (quantization step) changes spatially to optimize both compression and visual quality.
Data Source
AI summary
Multiview images may comprise attribute frames and geometry frames. Samples of a geometry frames may comprise depth information corresponding to collocated samples of the attribute frames. Additional edge feature frames may be generated, for the multiview images, with samples of the edge feature frame indicating whether collocated samples of the geometry frames are at edges and/or discontinuities. Information from the edge feature frame may be used to correct quantization errors that may be associated with samples, of the geometry frames, that are located at edges and discontinuities.