Patch-Based Image Synthesis Using Gradient Voting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional patch-based synthesis techniques are limited in handling geometric and photometric transformations, leading to poor performance in applications such as image completion, hole filling, and image stitching, especially when dealing with large intra-image variations and multiple source images.
Innovation Solution
The proposed method extends the patch search space to include geometric transformations like rotation, non-uniform scale, and reflection, and photometric transformations like gain and bias adjustments, integrating image gradients for improved matching and blending, allowing for more flexible and accurate synthesis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional patch-based synthesis techniques are used, then the method is simple and fast, but the performance is poor when handling geometric and photometric transformations
Solution Approach 1:
The patent makes the patch search space dynamic by incorporating geometric transformations (rotation, scaling, reflection) and photometric transformations (gain, bias adjustments) that adapt to the specific characteristics of source and target patches, allowing the system to handle diverse geometric and photometric variations effectively
Solution Approach 2:
The patent extends the traditional 2D patch search space by adding transformation parameters (rotation angles, scale factors, reflection flags, gain/bias values), creating a multi-dimensional search space that encompasses both spatial and photometric transformations, thereby significantly improving synthesis performance
2Manufacturing precision
If the patch search space is extended to include geometric and photometric transformations, then the synthesis accuracy improves, but the computational complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-defining the transformation space (geometric transformations like rotation and scaling, and photometric transformations like gain and bias) and pre-processing source patches with these transformations before the actual synthesis process, which streamlines the search and reduces computational overhead during execution
Solution Approach 2:
The patent systematically varies transformation parameters (rotation angles, scale factors, gain/bias values) to generate multiple transformed versions of source patches, allowing the synthesis algorithm to select the best matching patch while maintaining a balance between accuracy and computational efficiency through parameter optimization
3Reliability
If conventional color averaging is used, then the blending is simple and fast, but the handling of large intra-image variations is poor
Solution Approach 1:
The patent applies local quality by using gradient domain features to capture local variations in color and texture within patches, and by performing voting for colors and gradients at each pixel location, allowing the blending process to adapt to local intra-image variations rather than applying uniform averaging across the entire patch
4Manufacturing precision
If multiple source images are used, then the synthesis quality improves, but the difficulty of handling inconsistencies increases
Solution Approach 1:
The patent creates a universal framework that handles multiple source images by applying the same extended transformation space (geometric and photometric transformations) and gradient domain voting mechanism to all source images, providing a consistent and unified approach to handling inconsistencies across multiple sources
Data Source
AI summary
Methods, apparatus, and computer-readable storage media for patch-based image synthesis using color and color gradient voting. A patch matching technique provides an extended patch search space that encompasses geometric and photometric transformations, as well as color and color gradient domain features. The photometric transformations may include gain and bias. The patch-based image synthesis techniques may also integrate image color and color gradients into the patch representation and replace conventional color averaging with a technique that performs voting for colors and color gradients and then solves a screened Poisson equation based on values for colors and color gradients when blending patch(es) with a target image.