Avatar Mesh Rendering for Accurate Virtual Accessory Occlusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image generation systems fail to accurately represent depth when superimposing virtual objects on a player's body, leading to unnatural images due to improper occlusion and lack of unity between the player's body and virtual objects.
Innovation Solution
An image generation apparatus that recognizes a player's body, generates a mesh structure reflecting the player's skeleton, and superimposes virtual objects on this mesh structure based on acquired viewpoint information, accurately representing depth and occlusion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If virtual objects are superimposed on player body images, then entertainment and immersion are enhanced, but depth representation becomes inaccurate and images appear unnatural
Solution Approach 1:
The player's body is segmented into multiple mesh parts (head, torso, limbs, etc.) that can be independently positioned and rendered. This segmentation allows accurate depth calculation for each body part relative to virtual objects, enabling proper occlusion handling while maintaining immersive augmented reality experiences
Solution Approach 2:
A mesh structure serving as an intermediary model is introduced between the captured player image and the virtual objects. This mesh model provides accurate spatial information and depth relationships, enabling correct rendering of occlusion relationships without compromising the immersive quality of the augmented reality image
2Ease of manufacture
If virtual objects overwrite player body parts, then rendering simplicity is maintained, but depth accuracy and image naturalness deteriorate
Solution Approach 1:
The mesh structure of the player's body is generated and positioned in advance based on skeleton recognition results. This preliminary action establishes accurate depth information and spatial relationships before virtual objects are rendered, enabling proper occlusion handling without complicating the rendering process
Data Source
AI summary
An image generation apparatus 100 according to the present disclosure includes a player recognition section 102 configured to recognize a body of a player, a viewpoint acquisition section 103 configured to acquire viewpoint information including a viewpoint position and a viewpoint direction, a mesh generation section 105 configured to generate a mesh structure of the player that reflects a skeleton of the body of the player, on the basis of a result of the recognition, and an image generation section 110 configured to generate an image by rendering the mesh structure of the player and a virtual object when viewed in the viewpoint direction from the viewpoint position in the viewpoint information, and superimposing the rendered virtual object on the rendered mesh structure of the player.