Machine-learning for enhanced machine reading of non-ideal capture conditions
Inventors
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12626322-B1
Publication Date
2026-05-12
Expiration Date
Abstract
Implementations of the present disclosure include receiving a training image, providing a hash pattern that is representative of the training image, applying a plurality of filters to the training image to provide a respective plurality of filtered training images, identifying a filter to be associated with the hash pattern based on the plurality of filtered training images, and storing a mapping of the filter to the hash pattern within a set of mapping in a data store.
Core Innovation
The invention relates to remote deposit capture by receiving an image captured by a computing device and processing the image to generate an image hash pattern for the image. The processing includes scaling the image to generate a micro-image and applying an edge detection process to the micro-image. The edge detection process uses a convolutional neural network to generate a filtered micro-image in which the filtered micro-image includes one or more edges of one or more objects within the micro-image, and at least one of the one or more objects includes text.
From the filtered micro-image, the method generates the image hash pattern and filters the image using a first filter selected based on the image hash pattern to obtain a filtered image. Before filtering the image using the first filter, the method identifies the first filter based on a mapping of filters to hash patterns stored in a data store. The identification includes determining that the image hash pattern sufficiently matches a first hash pattern mapped to the first filter.
The sufficiency of the match is determined by calculating a Hamming distance between the image hash pattern and the first hash pattern and comparing the Hamming distance to a threshold Hamming distance. The invention then determines text data from the filtered image. The same overall operations are performed by a system and by a non-transitory computer-readable storage medium coupled to one or more processors.
Claims Coverage
The partial content includes three independent claims (a method, a system, and a non-transitory computer-readable storage medium), each covering the same core workflow with edge-detection-based image hash patterns and filter selection via a mapped hash-pattern comparison using Hamming distance to a threshold. The independent claims include four inventive elements.
Edge-detection-based image hash pattern generation using convolutional neural network
Processing an image by scaling to generate a micro-image, applying an edge detection process to generate a filtered micro-image using a convolutional neural network, where the filtered micro-image includes edges of objects within the micro-image and at least one object includes text, and generating the image hash pattern based on the filtered micro-image.
Filter selection via mapping of filters to hash patterns
Identifying a first filter based on a mapping of filters to hash patterns stored in a data store prior to filtering, where identifying the first filter includes determining that the image hash pattern sufficiently matches a first hash pattern mapped to the first filter.
Hash-match criterion using Hamming distance to a threshold
Determining that the image hash pattern sufficiently matches the first hash pattern mapped to the first filter by calculating a Hamming distance between the image hash pattern and the first hash pattern and comparing the Hamming distance to a threshold Hamming distance.
Text-data determination from the filtered image
After filtering the image using the first filter selected based on the image hash pattern to obtain a filtered image, determining text data from the filtered image.
Across the independent claims, the core coverage is directed to generating an image hash pattern from a filtered micro-image produced by convolutional neural-network edge detection, selecting a filter by matching the hash pattern to a stored mapping using Hamming distance compared to a threshold, filtering to obtain a filtered image, and determining text data from the filtered image.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?