A method for detecting text areas in a video image.

摘要

Method of detecting text fields in a video image, comprising the following steps: - calculating the horizontal gradient of the intensity of each pixel of said video image and adding, for each pixel of said video image, gradients horizontal pixels belonging to a horizontal window integration covering the pixel concerned so as to generate an aggregate gradients image (G), - transforming said aggregate gradients image (G) into a binary image (B) by operation thresholding consisting in comparing the value of the total gradient of each pixel of the aggregate gradients image with at least one threshold value, to assign each pixel either the value of gray level V1 representative of a pixel of text belongs to a text area or the gray level value V2 representative of a non-text pixel belongs to a text area, so that a im is obtained binary agen containing pixels of text with a value of gray level V1 and pixels of non-text with a level value V2 gray where pixels adjacent text defining text fields, and - applying line by line treatment morphology mathematics to said binary image in order to join the text zones distant horizontally over N pixels to form a single.