Image Processing Device Dynamic Depth Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in accurately converting 2D images into 3D images due to limitations in relative depth estimation and differences in display specifications, leading to depth estimation errors and user fatigue.
Innovation Solution
An image processing device that dynamically applies a depth estimation model in real-time and non-linearly changes the depth map based on the analysis of the content class and objects in each scene of the input 2D image, using on-device learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single depth estimation model is used for all 2D images, then the device complexity is low, but the depth estimation accuracy deteriorates for complex images with low correlation
Solution Approach 1:
The patent changes the parameter of depth estimation by selecting different models based on image characteristics. The processor analyzes image features (such as correlation, content type, or depth distribution) and dynamically switches between multiple depth estimation models to optimize accuracy for different image types, resolving the contradiction between maintaining low complexity and achieving high accuracy across diverse images.
Solution Approach 2:
The patent implements a dynamic model selection mechanism where the system adaptively chooses the appropriate depth estimation model based on real-time analysis of image characteristics. This dynamic approach allows the system to maintain high depth estimation accuracy for various image types without requiring a fixed complex multi-model architecture, as the selection logic itself manages the complexity.
2Reliability
If relative depth estimation is used without depth sensors, then the device cost is reduced, but the absolute depth estimation accuracy deteriorates
Solution Approach 1:
The patent transforms the depth estimation output by applying scaling and offset parameters derived from scene analysis or metadata. This parameter adjustment converts relative depth values into absolute depth estimates, allowing the system to maintain reliability using only camera-based relative depth estimation while improving absolute depth accuracy through mathematical transformation.
Solution Approach 2:
The patent introduces an intermediary processing layer that takes relative depth maps from monocular estimation and transforms them into absolute depth information using scene context, object size priors, or camera parameters as mediators. This intermediary step bridges the gap between relative and absolute depth measurement without requiring additional depth sensors.
3Ease of operation
If manual slide UI is provided for adjusting stereoscopic effect, then the ease of operation is improved, but the productivity of content delivery is reduced
Solution Approach 1:
The patent implements automatic stereoscopic effect adjustment where the system analyzes image content and autonomously optimizes depth map parameters and stereoscopic settings without requiring manual user intervention. This self-service approach maintains ease of operation by eliminating the need for users to manually adjust settings while preserving content delivery productivity through automated processing.
Solution Approach 2:
The patent employs feedback mechanisms where the system monitors user viewing conditions, device characteristics, and image content to automatically adjust stereoscopic parameters. This feedback-driven approach provides ease of operation through adaptive optimization while maintaining productivity by reducing manual adjustment requirements.
Data Source
AI summary
An image processing device for performing three-dimensional (3D) conversion and a method performed by the image processing device are provided. The image processing device includes a memory to store one or more instructions, and at least one processor configured to execute the one or more instructions stored in the memory. The processor may be configured to analyze a content class of an input image. The processor may be configured to obtain a depth estimation model corresponding to the content class of the input image in real time by using on-device learning, based on a result of analyzing the content class of the input image. The processor may be configured to obtain a depth map of the input image that reflects estimated depth information, based on the depth estimation model according to the on-device learning. The processor may be configured to perform 3D conversion for the input image, based on the depth map of the input image.