The processed table image is shown in the figure: (Baidu search image, just to try the processing method)
The main processing flow is:
1. Image preprocessing: binary image
2. Extract table information: horizontal and vertical lines and intersections
3. Remove table: use table information to remove the table in the figure
4. Morphologically process the remaining non-table information, and divide the text area into blocks
5. Finally, I used the recognition module I trained to rewrite the content on a blank image
The effect is not bad. . . . .