Compressing Deep Image Data Using Object-Based Depth Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep images containing depth information require significant memory and bandwidth, and existing compression methods often compromise editing capabilities, leading to undesirable artifacts.
Innovation Solution
A computer-implemented method for compressing image data with depth information by combining pixel samples based on object data and depth variations, using object identifiers to group samples from the same object and adjusting depth ranges to minimize errors, while allowing for easy editing of the compressed image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If deep image data is stored with full depth information for each pixel, then editing capabilities are maintained, but memory and bandwidth requirements increase significantly
Solution Approach 1:
The patent segments the pixel data into discrete samples with associated depth values and object identifiers. By organizing data into sample-level segments rather than storing complete per-pixel depth maps, the system reduces memory requirements while maintaining the ability to selectively edit individual objects or depth ranges.
Solution Approach 2:
The patent creates a compressed representation that copies only essential depth and object information rather than storing complete per-pixel depth maps. This allows the compressed image to retain sufficient information for editing operations while using significantly less memory than full deep image storage.
2Quantity of substance
If compression is applied to reduce memory usage, then storage efficiency improves, but editing capabilities are compromised introducing artifacts
Solution Approach 1:
The patent changes the parameter representation by storing depth values and object identifiers as discrete samples rather than continuous per-pixel data. This parameter transformation enables compression while preserving editing capabilities, as the sampled parameters can be selectively modified without affecting the entire image.
Solution Approach 2:
The compressed representation copies essential depth and object information at sample points rather than storing complete per-pixel depth maps. This selective copying maintains sufficient information for editing operations while achieving memory reduction, avoiding the artifacts that would result from aggressive compression of continuous data.
3Loss of information
If per-pixel depth information is stored for all objects, then complete scene information is preserved, but data volume increases
Solution Approach 1:
The patent segments scene information into discrete samples with associated depth values and object identifiers rather than storing complete per-pixel depth maps. This segmentation preserves essential scene information about object positions and depths while significantly reducing data volume through selective sampling.
Solution Approach 2:
The patent copies only essential scene information at sample points, including depth values and object identifiers, rather than storing complete per-pixel depth maps for all objects. This selective copying maintains sufficient scene information for editing and rendering while achieving substantial data volume reduction.
Data Source
AI summary
An image dataset is compressed by combining depth values from pixel depth arrays, wherein combining criteria are based on object data and/or depth variations of depth values in the first pixel image value array and generating a modified image dataset wherein a first pixel image value array represented in a received image dataset by the first number of image value array samples is in turn represented in the modified image dataset by a second number of compressed image value array samples with the second number being less than or equal to the first number.