Improved Method for Finding Subimages with Maximal Property
Image analysis and pattern recognition are rapidly developing and fruitful disciplines.
Evaluating images to locate a subimage with maximal property is a frequent operation.
For example, an object of a certain color is searched for within an image of many objects. This operation is currently very costly in terms of computing time and, therefore, computing power. There is a need for a method that improves upon existing methods with significantly increased computational speed.
Description of Technology
Currently, evaluating images to locate a subimage with maximal property requires a significant amount of computing power because it must consider and rank many
subimages. However, the present method considers only a select few subimages based on maximal lines of image pixels. This results in a significant reduction in computing power requirements and computing time. The method takes into consideration each one-dimensional row of a representative matrix and locates its maximal subrow, and then analyzes and ranks subimages within the boundaries of this maximal subrow. The resulting maximal subimage’s information is then recorded for comparison against subsequent rows’ resulting maximal subimages. After all rows of the representative matrix have been considered, the subimage with maximal property is obtained by comparing each rows’ maximal subimages against each other and selecting the one subimage therein with maximal property.
- Faster – six times faster at performing object and color detection
- artificial intelligence
- face detection
- pedestrian detection
- image processing
- multi-dimensional physical simulations
- computational simulations
Licensing Rights Available
Full licensing rights available
Inventors: Maxwell Reuter
Tech ID: TEC2016-0126
For Information, Contact:
Michigan State University