Image Blending via Gradient Parameter Maps for Natural Bokeh
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image capturing methods fail to produce full depth of field (DOF) images with clear main objects and natural bokeh effects, often resulting in discontinuous DOF or unnatural results, and are limited by the need for complex and time-consuming image capturing processes using large aperture lenses.
Innovation Solution
An image processing method that captures images with different focal lengths, performs geometric calibration, calculates pixel gradients, and generates parameter maps to blend images, producing clear main objects and blurry backgrounds, thereby achieving a natural bokeh effect or full DOF image without ghost phenomena.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a large aperture lens is used to achieve a better bokeh effect, then the bokeh effect is improved, but the device volume and cost increase
Solution Approach 1:
The patent changes the focal length parameter of the lens to achieve bokeh effect without requiring large aperture. By capturing images at multiple focal lengths and blending them, the system produces natural bokeh effects while maintaining a compact lens design suitable for consumer cameras.
Solution Approach 2:
The patent segments the image capture process into multiple shots at different focal lengths, then blends the results. This segmentation allows the system to achieve bokeh effect computationally rather than relying solely on optical properties of a large aperture lens.
2Manufacturing precision
If multiple images are captured and combined to produce full DOF image, then the full DOF is improved, but the image capturing time increases
Solution Approach 1:
The patent dynamically adjusts the focal length during a zooming operation to capture multiple images. By utilizing the natural zooming motion rather than requiring sequential focused shots, the system reduces capturing time while still obtaining multiple focal length images for blending.
Solution Approach 2:
The patent performs geometric calibration and feature point matching in advance to prepare for efficient image blending. This preliminary processing enables faster combination of multiple images while maintaining full DOF quality.
3Manufacturing precision
If images are blended to produce full DOF image, then the full DOF is improved, but ghost phenomena occur
Solution Approach 1:
The patent applies local quality by using gradient operations to analyze local image features and determine the in-focus region. By examining gradient magnitude and direction at each pixel location, the system accurately identifies which regions should be sharp versus blurred, preventing ghost phenomena while maintaining natural transitions.
Solution Approach 2:
The patent uses gradient comparison as feedback to guide the blending process. By comparing gradients between images and using them to weight the blending contribution, the system ensures that the final image maintains accurate focus regions without ghosting artifacts.
4Stability of the object's composition
If a fixed image capturing device is used to ensure geometric consistency, then the image stability is improved, but the operation complexity increases
Solution Approach 1:
The patent replaces the mechanical requirement of a fixed tripod with computational geometry correction. By detecting feature points and calculating geometric transformation parameters, the system automatically corrects for camera movement, eliminating the need for complex mechanical stabilization setups.
Data Source
AI summary
An image capturing device and an image processing method are provided. The present method includes following steps. A first image and a second image are captured with a first focal length and a second focal length correspondingly. The motion corrected second image is produced by performing geometric correction procedure on the second image. A gradient operation is performed on each of the pixels of the first image to obtain a plurality of first gradients, and the gradient operation is performed on each of the pixels of the motion corrected second image to obtain a plurality of second gradients. The first gradients and the second gradients are compared and a first parameter map is generated according to the comparison results. A blending image is produced in according with the first parameter map and the first image, and an output image is produced at least in according with the blending image.