Panning Image Generation via Layer Segmentation and Motion Stabilization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Generating panning images that effectively blur the background while maintaining the subject in focus is challenging due to the difficulty in precisely tracking the subject with a moving camera, often resulting in the subject being blurred as well.
Innovation Solution
A system that stabilizes reference and background images based on the motion of the image capture device, generates an object layer by segmenting the subject from the reference image, combines it with a blurred background layer to create a panning image, where the subject is kept sharp and the background is blurred, using directional blur kernels aligned with the camera motion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If the camera pans to follow the subject during capture, then the background is blurred to create a dramatic look, but the subject may become blurred due to imprecise tracking
Solution Approach 1:
The patent segments the image into multiple layers: a sharp object layer containing the subject and a blurred background layer. This segmentation allows the subject to remain in focus while the background is blurred, resolving the contradiction between achieving panning blur effect and maintaining subject focus precision.
Solution Approach 2:
The patent introduces motion information as an intermediary element that characterizes the camera's motion during capture. This motion information is used to stabilize images and generate directional blur kernels, serving as a mediator between the camera's actual motion and the final image processing to achieve precise subject focus while maintaining background blur.
2Area of stationary object
If the camera moves to capture more visual content, then the field of view increases, but the images become harder to stabilize and align
Solution Approach 1:
The patent performs preliminary stabilization by selecting a reference image and using motion information to stabilize other images relative to it. This preliminary action establishes a stable reference frame before generating the final panning image, enabling precise alignment even when the camera moves to capture a wider field of view.
Solution Approach 2:
The patent uses motion information as feedback to adjust and align images. The motion characteristics captured during the panning movement are fed back into the processing system to generate appropriate stabilization and blur effects, enabling precise image alignment while maintaining an expanded field of view.
3Quantity of substance
If multiple images are captured during panning motion, then more visual information is available, but determining which portions to present becomes difficult and time-consuming
Solution Approach 1:
The patent extracts and separates the object layer from the background layer, taking out only the relevant subject information while discarding unnecessary background variations. This extraction process automatically identifies and presents the most important visual content (the sharp subject) while filtering out redundant information, eliminating the need for manual selection and saving time.
Data Source
AI summary
Images may be captured by a moving image capture device. A reference image and a background image may be selected from the images. The reference image may include depiction of an object, with the object blocking view of the background. The background image may include depiction of the background blocked by the object in the reference image. An object layer may be generated by segmenting the depiction of the object from the reference image. A background layer may be generated by combining the depiction of the background in the background image with the reference image. The background layer may be blurred and combined with the object layer to generate a panning image.