Web Image Extraction Filtering by Annotation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for extracting image content related to business establishments from the web often provide images that do not accurately represent the establishments, due to inclusion of irrelevant or low-quality content.

Innovation Solution

A machine-implemented method that selects and filters image content from webpages based on annotations generated from characteristics such as obscenity probability, image type, and color, to provide high-quality, relevant images for display.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If image content is extracted from multiple webpages without filtering, then the quantity of images is increased, but the quality and relevance of images deteriorates

Engineering Contradiction:
Improvequantity of imagesVSAvoidquality and relevance of images
Core Design Contradiction:
Quantity of substanceVSManufacturing precision

Solution Approach 1:

The system performs preliminary actions by generating annotations for image content before filtering occurs. These annotations include characteristics such as image type, color properties, and obscenity probability assessments. By pre-characterizing the images through annotation, the system establishes a foundation for subsequent quality-based filtering that maintains both quantity and quality standards.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system implements feedback mechanisms by using generated annotations to inform the filtering process. The annotations provide feedback about image characteristics that guide the filtering algorithm to retain relevant, high-quality images while eliminating inappropriate or low-quality content. This feedback loop ensures that image quantity is maintained without sacrificing quality.

Inventive Principle:
Principle #23Feedback

2Loss of information

If all extracted image content is displayed, then the completeness of information is improved, but the appropriateness and relevance of displayed content deteriorates

Engineering Contradiction:
Improvecompleteness of image informationVSAvoidappropriateness and relevance of images
Core Design Contradiction:
Loss of informationVSReliability

Solution Approach 1:

The system extracts only the relevant and appropriate image content from the complete set of extracted images by applying filters based on generated annotations. The filtering process selectively removes inappropriate images (such as those with high obscenity probability) while retaining relevant business establishment images, thus maintaining information completeness without compromising appropriateness.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The system applies different quality standards to different images based on their annotated characteristics. Rather than applying a uniform filter, the system evaluates each image's specific properties (image type, color, obscenity probability) and makes localized decisions about which images to display. This ensures that the displayed image set maintains high overall appropriateness while preserving relevant information.

Inventive Principle:
Principle #3Local quality

3Manufacturing precision

If filtering based on multiple characteristics is applied, then the quality of displayed images is improved, but the complexity of the processing system increases

Engineering Contradiction:
Improvequality of displayed imagesVSAvoidcomplexity of filtering system
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The system segments the image filtering process into distinct, modular components: annotation generation (assessing image type, color, obscenity probability) and filtering based on annotated characteristics. By dividing the complex filtering task into separate processing stages, the system achieves high image quality through multi-characteristic evaluation while managing system complexity through modular architecture.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentUS9645981B1Extraction of business-relevant image content from the web
Publication Date: 2017.05.09 GOOGLE LLC
  • US9645981B1 patent drawing
  • US9645981B1 patent drawing
  • US9645981B1 patent drawing

AI summary

A system and machine-implemented method for providing image content corresponding to a business establishment is provided. Several webpages corresponding to a business establishment are received, and one or more webpages are selected from the several webpages, based on the content of each of the several webpages. At least one webpage related to the selected one or more webpages is retrieved. Image content is extracted from the retrieved at least one webpage. At least one annotation is generated for the extracted image content based on at least one characteristic of the extracted image content. The image content is filtered based on the generated at least one annotation for the extracted image content. The filtered image content is provided for display.