Depth Map Based Image Synthesis for Realistic Subject Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies fail to synthesize images with quality similar to photographs, particularly when generating images based on user-specified areas with subjects positioned along those areas.
Innovation Solution
An electronic device comprising a processor and memory, configured to identify images with colored areas, obtain depth maps based on these images, and generate virtual images with subjects indicated by the colors, using a method that includes identifying semantic maps, obtaining candidate depth maps, selecting a matching depth map, and synthesizing images based on the selected depth map and semantic map.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If existing image synthesis technologies are used, then images can be generated based on user-specified areas, but the quality of synthesized images is not similar to photographs
Solution Approach 1:
The patent segments the image synthesis process into multiple components: semantic map generation (identifying subject areas), depth map generation (determining spatial relationships), and final image synthesis. This segmentation allows each component to be optimized independently, with the depth map specifically addressing the realism issue by providing accurate spatial information that enhances photograph-like quality.
Solution Approach 2:
The patent introduces depth maps as an intermediary element between the semantic map and the final synthesized image. The depth map serves as a mediator that provides spatial relationship information, enabling more realistic subject positioning and perspective. This intermediary component bridges the gap between simple area-based synthesis and photograph-quality realism.
2Manufacturing precision
If depth maps are obtained based on semantic maps using multiple candidate depth maps, then image realism is enhanced, but processing time and computational complexity increase
Solution Approach 1:
The patent generates multiple candidate depth maps (excessive action) but then selects only the most appropriate one or combines them selectively (partial action). This approach ensures high positioning accuracy by considering multiple possibilities while avoiding the full computational cost of processing and integrating all candidate depth maps equally, thus reducing overall processing time.
3Manufacturing precision
If multiple candidate depth maps are generated and selected, then image quality improves, but device complexity increases
Solution Approach 1:
The patent performs preliminary generation of multiple candidate depth maps based on the semantic map before the final synthesis step. This preliminary action allows the system to pre-compute and evaluate different depth possibilities, selecting the most accurate one for the final image. This approach improves depth map accuracy while managing device complexity by organizing the process into distinct phases rather than requiring all candidates to be processed simultaneously.
Data Source
AI summary
An electronic device according to one embodiment may include memory storing instructions and at least one processor operably coupled to the memory. The at least one processor may be configured to, when the instructions are executed, identify a first image comprising one or more areas distinguished by one or more colors; obtain at least one depth map based on the first image, wherein the at least one depth map comprises the one or more areas in the first image; and obtain, based on the first image and the at least one depth map, a virtual image including one or more subjects indicated by colors of the one or more areas.