Systems and methods for mobile image capture and content processing of driver's licenses

A method comprising using at least one hardware processor to:

• receive an image of a document captured by a mobile device;
• by each of two or more different image-cropping methods, determine coordinates of proposed corners of the document in the image, wherein the two or more different image-cropping methods include feature matching, and wherein the feature matching comprises
• deriving an array of features from the image,
• identifying one or more templates of the document by finding a correspondence between the array of features and pre-calculated features associated with the one or more templates, and
• estimating four corners of the document in the image based on the identified one or more templates;
• determine cropping parameters based on the coordinates of the proposed corners of the document determined by the two or more different image-cropping methods;
• crop the image using the determined cropping parameters;
• process the cropped image according to one or more image processing methods;
• and
• perform optical character recognition on the processed cropped image.

The method comprising using at least one hardware processor to:

• receive an image of a document captured by a mobile device;
• by each of two or more different image-cropping methods, determine coordinates of proposed corners of the document in the image;
• determine cropping parameters based on the coordinates of the proposed corners of the document determined by the two or more different image-cropping methods;
• crop the image using the determined cropping parameters;
• process the cropped image according to one or more image processing methods, wherein processing the cropped image comprises
• identifying a photograph in the cropped image,
• determining an aspect ratio of the photograph in the cropped image, and
• identifying a size of the document based on the determined aspect ratio;
• and
• perform optical character recognition on the processed cropped image.

A system comprising:

• at least one hardware processor;
• a non-transitory computer-readable medium storing software;
• and
• the software configured to, when executed by the at least one hardware processor,
• receive an image of a document captured by a mobile device,
• by each of two or more different image-cropping methods, determine coordinates of proposed corners of the document in the image, wherein the two more different image-cropping methods include one or both of rounded corner detection or feature matching,
• wherein the rounded corner detection comprises detecting edges in the image, determining connected points using contour tracing on the detected edges, detecting splines by scanning connected points in a local window, and identifying rounded corners from the detected splines, and
• wherein the feature matching comprises deriving an array of features from the image, identifying one or more templates of the document by finding a correspondence between the array of features and pre-calculated features associated with the one or more templates, and estimating four corners of the document in the image based on the identified one or more templates,
• determine cropping parameters based on the coordinates of the proposed corners of the document determined by the two or more different image-cropping methods,
• crop the image using the determined cropping parameters,
• process the cropped image according to one or more image processing methods, and
• perform optical character recognition on the processed cropped image.

A non-transitory computer-readable medium having instructions stored thereon, wherein the instructions, when executed by a processor, cause the processor to:

• receive an image of a document captured by a mobile device;
• by each of two or more different image-cropping methods, determine coordinates of proposed corners of the document in the image, wherein the two more different image-cropping methods include one or both of rounded corner detection or feature matching,
• wherein the rounded corner detection comprises detecting edges in the image, determining connected points using contour tracing on the detected edges, detecting splines by scanning connected points in a local window, and identifying rounded corners from the detected splines, and
• wherein the feature matching comprises deriving an array of features from the image, identifying one or more templates of the document by finding a correspondence between the array of features and pre-calculated features associated with the one or more templates, and estimating four corners of the document in the image based on the identified one or more templates;
• determine cropping parameters based on the coordinates of the proposed corners of the document determined by the two or more different image-cropping methods;
• crop the image using the determined cropping parameters;
• process the cropped image according to one or more image processing methods;
• and
• perform optical character recognition on the processed cropped image.