Image Processing With Object Tracking for Artifact Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques for removing unwanted objects in digital photography are computationally expensive and may result in sub-optimal capture timing, leading to unwanted artifacts due to scene changes.
Innovation Solution
An apparatus and method for tracking unwanted objects across multiple images, estimating performance metrics, and optimizing the fill-in processing operation based on these metrics to reduce computational burden and improve timing for optimal image capture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If fill-in processing is performed on all subsequently-received images, then complete removal of unwanted objects is achieved, but computational cost and processing time increase significantly
Solution Approach 1:
The patent extracts and identifies only the most suitable reference image from the sequence of subsequently-received images based on performance metrics, rather than processing all images. This selective extraction approach reduces computational burden while maintaining effective fill-in processing.
Solution Approach 2:
The patent introduces performance metrics as parameters to evaluate and select the optimal reference image. By changing from processing all images to processing only the image with optimal metrics, the system achieves efficiency improvement while maintaining processing quality.
2Reliability
If fill-in processing is performed on all subsequently-received images, then object removal completeness is maintained, but processing time increases
Solution Approach 1:
The system extracts only the optimal reference image based on evaluated performance metrics, eliminating the need to process all subsequent images. This extraction strategy maintains object removal completeness while significantly reducing processing time.
Solution Approach 2:
The patent performs preliminary evaluation of performance metrics on subsequently-received images before executing fill-in processing. This preliminary action identifies the optimal reference image in advance, preventing unnecessary processing of other images and reducing overall processing time.
3Reliability
If fill-in processing is performed on images with sub-optimal timing, then object removal is achieved, but unwanted artifacts appear
Solution Approach 1:
The patent implements feedback through performance metrics that evaluate the suitability of subsequently-received images for fill-in processing. This feedback mechanism guides the selection of the optimal reference image, ensuring that processing is performed only when conditions are favorable, thereby minimizing artifacts and maintaining image quality.
Solution Approach 2:
The system uses performance metrics as parameters to determine the optimal timing for fill-in processing. By monitoring these parameters across multiple images, the system identifies the moment when processing will yield the best results, thus avoiding artifacts caused by sub-optimal timing.
4Reliability
If fill-in processing is performed on multiple images, then object removal completeness improves, but energy consumption increases
Solution Approach 1:
The patent extracts and selects only the single most suitable reference image from multiple subsequently-received images based on performance metrics. This extraction approach maintains object removal completeness while avoiding the energy consumption associated with processing multiple images.
Solution Approach 2:
The system introduces performance metrics as energy-efficient parameters for selecting the optimal reference image. By using these parameters to guide processing, the system achieves complete object removal while minimizing energy consumption compared to processing all subsequent images.
Data Source
AI summary
An apparatus and method for image processing is disclosed. The method may include receiving an image from a camera sensor, receiving selection of one or more target objects appearing in the image and tracking the one or more target objects over a plurality of subsequently-received images. For the subsequently-received images in turn, the method may include estimating one or more performance metric(s) associated with performing a fill-in processing operation of the one or more tracked target objects and saving the image as an optimised reference image if the respective performance metric(s) indicate an improved performance over that of one or more previously-received images from the time of receiving selection. The method may include performing the fill-in processing operation using one or more of the saved optimised reference images for output to a display screen.