Deep Image Identifiers for Hierarchical Object Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In computer graphics, deep image compositing requires selecting regions of a 2-D image based on complex conditions spanning different domains, which is time-consuming and prone to repetition due to the loss of hierarchical and structural information during the rendering process.
Innovation Solution
Assigning deep identifiers to geometric primitives in a 3-D virtual scene before rendering, which encode hierarchical and structural information, allowing for flexible selection and manipulation of objects in post-rendering processes by propagating these identifiers through the rendering and compositing operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If hierarchical and structural information is retained through rendering, then object selection and manipulation in post-rendering becomes efficient and flexible, but the rendering process complexity increases due to identifier generation and propagation
Solution Approach 1:
The patent applies preliminary action by generating deep identifiers for geometric primitives before the rendering process. These identifiers encode hierarchical and structural information about objects in the 3-D scene, allowing post-rendering operations to efficiently select and manipulate objects without requiring complex queries or additional rendering passes. The identifier generation and assignment occurs in advance, preparing the data structure for efficient later operations.
2Loss of information
If deep identifiers are propagated through rendering operations, then post-rendering operations can efficiently select and manipulate objects, but information loss during rendering is prevented which increases data processing requirements
Solution Approach 1:
The patent implements copying by creating deep identifiers that are propagated through the rendering process to associate rendered samples with their source geometric primitives. Instead of retaining complex object data through rendering, the system creates compact identifier copies that encode hierarchical relationships. These identifier copies are attached to rendered samples, enabling efficient object selection and manipulation in post-rendering without requiring the original complex scene graph data.
3Adaptability or versatility
If multiple samples per pixel are maintained in deep image format, then advanced compositing operations can be performed, but the image data size and processing complexity increase
Solution Approach 1:
The patent uses copying to associate deep identifiers with individual samples in deep images. Each sample in the multi-sample per pixel structure receives an identifier copy that links it to the source geometric primitive. This allows compositing operations to efficiently select and manipulate specific objects within the deep image by querying their identifiers, without requiring complex spatial or hierarchical analysis of the sample data structure.
Data Source
AI summary
A method may include receiving a plurality of objects from a 3-D virtual scene. The plurality of objects may be arranged in a hierarchy. The method may also include generating a plurality of identifiers for the plurality of objects. The plurality of identifiers may include a first identifier for a first object in the plurality of objects, and the identifier may be generated based on a position of the first object in the hierarchy. The method may additionally include performing a rendering operation on the plurality of objects to generate a deep image. The deep image may include a plurality of samples that correspond to the first object. The method may further include propagating the plurality of identifiers through the rendering operation such that each of the plurality of samples in the deep image that correspond to the first object are associated with the identifier.