Paint-Based Image Search for Irregular Object Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image search technologies face challenges in accurately identifying search targets within images, especially when the target is not rectangular or consists of multiple objects, as they struggle to distinguish between foreground and background, leading to unsatisfactory search results.
Innovation Solution
The method involves a user device that allows users to paint the search target on an input image, generating region data, and then preprocesses this data to eliminate background and extract features, transforming it for an image search, enabling the system to distinguish between multiple objects and separate them from the background.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a rectangular region is used to specify the search target, then the operation is simple, but the search precision deteriorates when the target is not rectangular or consists of multiple objects
Solution Approach 1:
The patent segments the image into multiple regions and allows users to select multiple discrete objects independently. Instead of forcing a single rectangular selection, the system divides the image space into selectable regions that can be individually chosen, enabling precise specification of multiple objects or irregularly shaped targets while maintaining ease of operation through intuitive region-based selection.
2Quantity of substance
If a large rectangle is used to cover multiple objects, then all objects are included, but the search precision deteriorates due to inclusion of background
Solution Approach 1:
The patent extracts and removes the background from the selected regions. After users select multiple objects or regions of interest, the system identifies and separates the foreground objects from the background, eliminating background interference from the search query. This extraction process ensures that only the relevant objects are used for image search, maintaining precision even when multiple objects are selected.
3Measurement precision
If a small rectangle is used to specify one object from multiple objects, then the search precision for that object improves, but the ease of operation deteriorates as users must repeatedly adjust the rectangle
Solution Approach 1:
The patent performs preliminary segmentation of the image into multiple selectable regions before the user makes selections. By pre-dividing the image into meaningful regions (such as individual objects or semantic segments), the system eliminates the need for users to repeatedly adjust rectangles to isolate each object. Users can directly select from the pre-segmented regions, maintaining precision while significantly improving operational convenience.
Data Source
AI summary
Technologies generally described herein relate to a computing device for an image search. Example computing devices may include at least one processor and at least one memory storing instructions. The at least one processor may execute the instructions to perform operations. The operations may include obtaining an input image including at least one object that is a target for a search; displaying the input image on a display screen associated with the computing device; receiving an operation to paint the at least one object in the displayed input image to generate region data for at least one painted region; and displaying, on the display screen, one or more search results by an image search conducted based, at least in part, on the input image and the region data.