Silhouette Image Generation Using Composite Difference and Smoothness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing silhouette image generation methods, such as background subtraction, fail to produce accurate images when targets have shadows, blown out highlights, or are transparent, as they struggle to distinguish foreground from background.
Innovation Solution
An information processing device generates a silhouette image by calculating differences and smoothness between target and standard images across various imaging conditions, using a composite difference and composite smoothness to create a robust silhouette image without user input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If background subtraction method is used to generate silhouette image, then the method is simple and fast, but the silhouette image becomes inaccurate when targets have shadows, blown out highlights, or are transparent
Solution Approach 1:
The patent segments the silhouette image generation process into multiple independent processing stages: shadow detection and removal, highlight detection and removal, transparent object detection and handling, and final silhouette extraction. Each stage processes specific aspects of the image independently, allowing complex preprocessing while maintaining overall efficiency.
Solution Approach 2:
The patent introduces intermediate processing steps between the original image and the final silhouette image. These intermediaries include shadow-masked images, highlight-masked images, and processed images that gradually transform the input toward the final output, enabling precise handling of difficult cases without excessive complexity.
2Measurement precision
If multiple processing steps are added to improve silhouette image accuracy, then the accuracy improves, but the processing complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-detecting and marking shadow regions, highlight regions, and transparent object regions before the main silhouette extraction process. This preliminary segmentation allows the main processing to work with pre-processed data, reducing the complexity of the core algorithm while maintaining high accuracy.
Solution Approach 2:
The patent changes processing parameters dynamically based on detected image conditions. For example, different processing strategies are applied depending on whether shadow regions, highlight regions, or transparent objects are detected, allowing the system to adapt to various scenarios without requiring complex universal algorithms.
3Measurement precision
If user-guided input is required for silhouette image generation, then the accuracy improves, but the ease of operation decreases
Solution Approach 1:
The patent implements self-service by enabling the system to automatically detect and handle shadows, highlights, and transparent objects without user intervention. The algorithm autonomously identifies these challenging regions and applies appropriate processing, making the system as accurate as user-guided methods while maintaining ease of operation.
Solution Approach 2:
The patent incorporates feedback mechanisms where the system continuously evaluates the quality of detected regions and adjusts processing parameters accordingly. This automated feedback loop ensures high accuracy by adapting to the specific characteristics of each image without requiring user input to guide the process.
Data Source
AI summary
The disclosure is an information processing device including: a memory; and at least one processor coupled to the memory and performing operations. The operations includes: generating, based on a target image including a target and a standard image not including the target, a target structure image indicating a shape feature of an object included in the target image and a standard structure image indicating a shape feature of an object included in the standard image, in each of a plurality of imaging conditions; calculating an individual difference being a difference between the target structure image and the standard structure image, and a composite difference based on the individual differences; calculating an individual smoothness being a smoothness of the target structure image, and a composite smoothness based on the individual smoothness; and generating a silhouette image of the target, based on the composite difference and the composite smoothness.