Backward Depth Mapping for Stereoscopic Image Synthesis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing backward depth mapping methods for stereoscopic image synthesis result in pixel gaps and cracks in the virtual view due to rounding of horizontal and vertical coordinates and disparities, leading to incorrect mapping and foreground-background content overlap.

Innovation Solution

The backward depth mapping method calculates a second coordinate and depth value for the reference view based on a predetermined virtual view coordinate and depth value, using an error value algorithm to determine an actual depth value that minimizes pixel gaps and cracks, ensuring accurate mapping without foreground-background overlap.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If forward depth mapping is used to map pixels from reference view to virtual view, then the mapping process is simple and direct, but pixel gaps and cracks are generated in the virtual view due to rounding of coordinates and disparities

Engineering Contradiction:

Improvemapping process simplicityVSAvoiddepth map accuracy

Core Design Contradiction:

Ease of manufactureVSManufacturing precision

Solution Approach 1:

The patent inverts the traditional forward mapping approach by implementing backward mapping. Instead of mapping from reference view to virtual view, it maps from virtual view to reference view. This inversion allows each pixel in the virtual view to be the source of the mapping operation, ensuring that all pixels in the virtual view are properly covered while minimizing gaps and cracks caused by coordinate rounding.

Inventive Principle:

Principle #13The other way round (Inversion)

2Area of stationary object

If every pixel of the reference view is mapped to the virtual view, then the mapping coverage is complete, but coordinate rounding causes pixel gaps and cracks between pixels in the virtual view

Engineering Contradiction:

Improvemapping coverageVSAvoidpixel continuity

Core Design Contradiction:

Area of stationary objectVSManufacturing precision

Solution Approach 1:

The patent reverses the mapping direction to start from virtual view pixels rather than reference view pixels. This ensures that every pixel in the virtual view is accounted for as a source, and the corresponding reference view pixels are identified through inverse projection, eliminating gaps and cracks in the virtual view depth map.

Inventive Principle:

Principle #13The other way round (Inversion)

Solution Approach 2:

The patent introduces a new dimension to the mapping process by considering the virtual view as the primary reference frame. By changing the reference dimension from reference view coordinates to virtual view coordinates, the mapping process ensures complete coverage without gaps, as each virtual view pixel becomes the origin of its own mapping operation.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

3Device complexity

If disparity values are rounded to integer coordinates, then the computational complexity is reduced, but foreground-background content overlap occurs due to incorrect mapping

Engineering Contradiction:

Improvecomputational complexityVSAvoidmapping accuracy

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent inverts the mapping approach to work from virtual view to reference view, which maintains computational simplicity while improving accuracy. By starting with virtual view pixels and finding their corresponding reference view positions through inverse projection, the method avoids foreground-background overlap while keeping computational complexity manageable through integer coordinate operations.

Inventive Principle:

Principle #13The other way round (Inversion)

4Area of stationary object

If median filter is applied to fill pixel gaps in the depth map, then the gaps are filled, but the depth edges of the virtual view are damaged

Engineering Contradiction:

Improvepixel gap fillingVSAvoiddepth edge integrity

Core Design Contradiction:

Area of stationary objectVSManufacturing precision

Solution Approach 1:

The patent applies preliminary action by preventing pixel gaps and cracks from occurring in the first place through backward mapping. By inverting the mapping direction to start from virtual view pixels, the method ensures complete coverage before any filtering is needed, thereby eliminating the need for median filtering that would damage depth edges.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent converts the potential harm of coordinate rounding into a benefit by inverting the mapping approach. Instead of rounding causing gaps in forward mapping, the backward mapping approach uses the same rounding operations to ensure that every virtual view pixel is properly covered, turning the rounding error source into a coverage guarantee.

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

Data Source

PatentUS10110873B2Backward depth mapping method for stereoscopic image synthesis

Publication Date: 2018.10.23 NAT CHIAO TUNG UNIV

AI summary

A backward depth mapping method for stereoscopic image synthesis includes: providing a first coordinate and a first depth value of a pixel of a virtual view in a stereoscopic image; calculating a second coordinate and a second depth value of a pixel of reference view, which is obtained by mapping the virtual view, based on the first coordinate and the first depth value of the pixel of the virtual view; calculating an error value corresponding to the first depth value based on the second coordinate and second depth value; and determining a third depth value of the pixel of the virtual view based on the error value. As a result, it is possible to avoid the virtual view producing a crack and the pixel of the background from appearing at the crack of the foreground.