Abstract: An example method of processing images of deformed indicia-bearing surfaces includes: detecting, within a document image, a plurality of image fragments, wherein each image fragment of the plurality of image fragment contains a respective sequence of alphabet symbols; grouping the plurality of image fragments by lines of text to be reconstructed in the document image; generating a map of isolines associated with the document image, wherein an isoline identifies a set of points that lie on a straight line of an undistorted image corresponding to the document image; generating a reverse transformation matrix that defines a set of transformations to be applied to the document image in order to remove image distortions caused by deformations of an indicia bearing surface; and generating an undistorted document image by applying the reverse transformation matrix to the document image.

Abstract: An example method of processing images of deformed indicia-bearing surfaces includes: detecting, within a document image, a plurality of image fragments, wherein each image fragment of the plurality of image fragment contains a respective sequence of alphabet symbols; grouping the plurality of image fragments by lines of text to be reconstructed in the document image; generating a map of isolines associated with the document image, wherein an isoline identifies a set of points that lie on a straight line of an undistorted image corresponding to the document image; generating a reverse transformation matrix that defines a set of transformations to be applied to the document image in order to remove image distortions caused by deformations of an indicia bearing surface; and generating an undistorted document image by applying the reverse transformation matrix to the document image.

Abstract: Aspects and implementations provide for techniques of classifying images by source types for efficient, fast, and economical processing of such images. The disclosed techniques include, for example, obtaining an input into an image processing operation (IPO input). The techniques further include processing, using a first neural network (NN), a first image associated with the IPO input to obtain a first feature vector, and processing, using a second NN, a plurality of second images associated with the IPO input to obtain a second feature vector. The techniques further include identifying, using the first feature vector and the second feature vector, a type of source used to generate the IPO input.