| 摘要 |
The invention concerns a method for detecting text in a video image. It comprises the following steps: calculating the horizontal gradient of the intensity of each pixel of the video image and adding, for each pixel of the video image, the horizontal gradients of the pixels belonging to a horizontal integration window including the pixel concerned to generate a sum image of gradients, transforming said sum image of gradients into a binary image comprising text pixels having a grey level V1 and non-text pixels having a grey level V2, the neighbouring text pixels being assembled in the text zones, and applying, line by line, a mathematical morphological processing to the binary image so as to connect the text zones horizontally distant by at most N pixels. |
