| CPC G06V 30/153 (2022.01) [G06V 30/1478 (2022.01); G06V 30/18105 (2022.01); G06V 30/242 (2022.01); G06V 30/18095 (2022.01); G06V 30/19107 (2022.01)] | 5 Claims |
|
4. An information processing method comprising:
obtaining, with at least one processor operating with a memory device in a computer, a character string image which includes a plurality of characters, and which includes the plurality of characters arranged in an arrangement direction;
acquiring, with the at least one processor operating with the memory device in the computer, a probability image representing a probability of an existence of a character in each pixel included in the character string image;
wherein the probability image is acquired using a machine learning model which outputs a region score image and an affinity score image;
obtaining, with the at least one processor operating with the memory device in the computer, a plurality of character regions in which the plurality of characters are estimated to respectively exist in the character string image based on the acquired probability image;
obtaining, with the at least one processor operating with the memory device in the computer, an additional character region which is located in the character string image after obtaining the plurality of character regions based on the probability image, and which does not overlap the plurality of obtained character regions based on a determination result on whether or not a pixel of a non-background color exists in a direction perpendicular to the arrangement direction at every position on the arrangement direction in the character string image;
recognizing, with the at least one processor operating with the memory device in the computer, the plurality of characters from the plurality of obtained character regions and the additional character region;
wherein the arrangement direction is an x direction or a y direction in the character string image,
determining, in the obtaining the additional character region, whether or not each of a plurality of columns which has a plurality of pixels arranged in the direction perpendicular to the arrangement direction is a candidate column including a pixel having the non-background color, the plurality of columns arranged in the arrangement direction in the character string image, and
wherein a region which corresponds to a range where the candidate columns continuously exist in the character string image, and which does not overlap the plurality of obtained character regions are obtained as the additional character region.
|