Depth Data Dilation for 3D Video Compression Artefact Reduction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Compression of 3D video data using quantized DCT coefficients can result in artificial irregular local displacements of object depth boundaries, leading to visible jagged edges when the viewpoint is changed, as the decompressed depth data no longer coincides with intensity edges, causing artefacts.

Innovation Solution

Performing dilation of depth data before compression, where depth edges are moved into regions with greater distance, and adaptively setting the amount of dilation based on the coincidence of detected positions and gradients in the image to minimize artefacts, thereby preventing errors during compression and decompression.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If depth data is encoded using quantized DCT coefficients for compression, then compression efficiency is improved, but depth boundary accuracy deteriorates causing visible jagged edges

Engineering Contradiction:

Improvecompression efficiencyVSAvoiddepth boundary accuracy

Core Design Contradiction:

ProductivityVSManufacturing precision

Solution Approach 1:

The patent applies dilation to the depth data before compression encoding. This preliminary action modifies the depth boundaries in advance so that when compression artifacts occur during quantized DCT encoding, the displaced depth edges still fall within the dilated regions, preventing visible jagged edges while maintaining compression efficiency.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The dilation operation creates a buffer zone around depth boundaries before compression. This cushioning effect ensures that even if compression causes local displacements of depth edges, the artifacts remain hidden within the dilated regions, thus protecting against visible artifacts while maintaining high compression ratios.

Inventive Principle:

Principle #11Beforehand cushioning (Prior cushioning)

2Manufacturing precision

If depth edges are preserved exactly at intensity edges, then depth accuracy is improved, but compression artifacts increase causing visible jagged edges

Engineering Contradiction:

Improvedepth edge precisionVSAvoidcompression artifacts

Core Design Contradiction:

Manufacturing precisionVSObject-affected harmful factors

Solution Approach 1:

The patent converts the harmful effect of compression-induced depth edge displacement into a beneficial effect by pre-dilating the depth data. The displacement that would normally create artifacts now occurs within the dilated regions, transforming what would be a defect into a feature that prevents visible artifacts while maintaining acceptable depth precision.

Inventive Principle:

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

3Object-affected harmful factors

If dilation is applied to depth data before compression, then artifact reduction is improved, but processing complexity increases

Engineering Contradiction:

Improveartifact reductionVSAvoidprocessing complexity

Core Design Contradiction:

Object-affected harmful factorsVSDevice complexity

Solution Approach 1:

The patent modifies the depth data by changing its spatial parameters through dilation before compression. This parameter change involves expanding the depth boundaries by a controlled amount, which prevents artifacts while adding only minimal processing complexity compared to the overall compression and rendering pipeline.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentEP2106668B2Image processing system for processing combined image data and depth data

Publication Date: 2022.01.19 KONINKLIJKE PHILIPS NV

Figure 1
Figure 2
Figure 3~4

AI summary

A combination of image data and depth data is supplied defining luminance and/or color for image positions in an image and distance to objects visible at said image positions respectively. The image data and the depth data are compressed by a compressor (120) and decompressed by a decompressor (16). The depth data is dilated before said compressing, to move an edge between a region of image positions with a relatively smaller distance and a region of image positions with a relatively greater distance towards the region of image positions with a relatively greater distance.