Image Synthesis Apparatus Frequency Component Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Poisson blending techniques often result in pixel value distortion due to low integrability of gradients, leading to unintended color artifacts and noticeable boundaries between synthesized images.
Innovation Solution
An image synthesis apparatus that calculates low-frequency and high-frequency components from multiple resolution images, allowing for the generation of a composite image that adjusts luminance and hue based on the background while reflecting the texture of the foreground, thereby minimizing pixel value distortion and making boundaries unnoticeable.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If Poisson blending is used to minimize gradient error between two images, then the boundary between images becomes less noticeable, but pixel value distortion occurs due to low integrability of the gradient field
Solution Approach 1:
The patent segments the image synthesis process into two distinct stages: first synthesizing the low-frequency component (luminance and hue) using Poisson blending, then synthesizing the high-frequency component (texture) separately and combining them. This segmentation allows each stage to focus on specific frequency ranges, reducing the pixel value distortion that occurs when attempting to optimize all frequencies simultaneously in traditional Poisson blending.
Solution Approach 2:
The patent changes the parameter being optimized from the full gradient field to specifically the low-frequency gradient field during the first synthesis stage. By adjusting the frequency parameter and synthesizing only low-frequency components initially, the method maintains integrability and reduces pixel value distortion while still achieving smooth boundary blending. The high-frequency components are then added in a subsequent stage without the same distortion issues.
2Object-affected harmful factors
If traditional Poisson blending optimizes the gradient field, then boundary visibility is reduced, but unintended color artifacts are generated due to mathematical constraints
Solution Approach 1:
The patent separates luminance/hue synthesis from texture synthesis by processing low-frequency and high-frequency components in distinct stages. The low-frequency Poisson blending handles luminance and hue transitions smoothly without generating color artifacts, while the high-frequency component is synthesized separately to preserve texture details without introducing the unintended color artifacts that plague traditional single-stage Poisson blending.
Solution Approach 2:
The patent introduces an intermediary processing step involving frequency decomposition and separate synthesis pathways. By using low-frequency and high-frequency component decomposition as intermediaries, the system can apply Poisson blending selectively to the low-frequency channel where it is most effective, while handling the high-frequency channel differently to avoid artifact generation.
3Manufacturing precision
If frequency component decomposition is used to separate luminance/hue and texture information, then image synthesis quality improves, but computational complexity increases
Solution Approach 1:
The patent segments the computational process into distinct frequency-based stages: low-frequency component extraction and synthesis, followed by high-frequency component extraction and synthesis. While this segmentation improves image synthesis quality by treating different frequency components appropriately, it does increase computational complexity compared to single-stage methods. The patent accepts this trade-off as the quality improvement in boundary blending and artifact reduction outweighs the additional computational requirements.
Data Source
AI summary
An image synthesis apparatus for generating a composite image by synthesizing a second image with a first image, includes an image acquisition unit configured to acquire first group images including the first image and a plurality of different resolution images corresponding to the first image, and second group images including the second image and a plurality of different resolution images corresponding to the second image, a frequency component calculation unit configured to calculate low-frequency components and high-frequency components from the first group images and the second group images, and an image synthesis unit configured to generate the composite image on the basis of the low-frequency components of the first group images and the high-frequency components of the second group images.