Viewport Selection Using Feature Weight Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in automatically selecting the most interesting or appealing sub-portions of panoramic images for display, as they struggle to effectively identify and serve salient regions without distorting the image.
Innovation Solution
A computer-implemented method and system that uses a weight vector with feature weights to score and select viewports within an image, trained using panoramic images with identified most interesting and least interesting viewports, to determine the most salient sub-portions for display, considering aspect ratios and overlap, and provides these selected viewports along with search results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If automatic viewport selection is implemented, then user experience is improved by displaying interesting regions, but system complexity increases due to feature extraction and scoring mechanisms
Solution Approach 1:
The system performs preliminary actions by pre-processing images to detect features and pre-calculating viewport scores during off-peak times or during image upload. This prepares the data in advance, so when a user requests an image, the viewport selection can be performed quickly without adding perceived complexity to the user experience.
Solution Approach 2:
The patent introduces intermediary components such as feature detectors, score calculators, and selection algorithms that act as mediators between the raw image data and the final viewport selection. These intermediaries break down the complex task of viewport selection into manageable stages, making the system more tractable while still delivering improved user experience.
2Measurement precision
If multiple features are analyzed for viewport scoring, then selection accuracy is improved, but processing time increases
Solution Approach 1:
The patent segments the viewport selection process into distinct stages: feature detection, feature extraction, scoring calculation, and viewport selection. Each stage processes specific aspects of the image independently, allowing for optimized processing at each step and enabling parallel computation where possible, thus reducing overall processing time while maintaining selection accuracy.
Solution Approach 2:
The system dynamically adjusts parameters such as the number of features to analyze, the complexity of feature extraction, and the scoring algorithm based on image characteristics and user preferences. This allows the system to maintain high selection accuracy when needed while reducing processing time for less critical images or when resources are constrained.
3Measurement precision
If training data is used to tune feature weights, then viewport selection accuracy is improved, but data processing requirements increase
Solution Approach 1:
The patent implements partial training by using a subset of available training data or by training on selected features only, rather than processing all possible training data and features. This approach achieves sufficient viewport selection accuracy without the prohibitive computational costs of exhaustive training, balancing accuracy improvements with data processing requirements.
Data Source
AI summary
Aspects of the disclosure provide for selecting viewports. A viewport is a sub-portion of an image. For example, a request to select a viewport for a particular image may be received. A set of potential viewports are identified within the particular image, and features are identified within the viewports of the set of potential viewports. A score is calculated for each given viewport of the set of potential viewports based at least in part on a weight vector having a set of feature weights and the identified features within that given viewport of the set of potential viewports. A viewport of the set of potential viewports is selected based on the calculated scores. The selected viewport is then provided in response to the request to select the viewport for the particular image.