How to get a line from image?

I try to redraw a grid from the OCR table.So how to get starting and ending point of a line?

I'm not sure what you are trying to do exactly. How are you trying to redraw a line in an image? Are you trying to do this on the image before recognizing it or after? Please give me more details.

After recognizing,I try to make a new grid in my image same as the one in source OCR document.

I'm afraid I sill don't know what you're trying to do.
After recognizing, you will have recognition results. If you save them to disk or memory, you will have some sort of document file, but you will probably not be still dealing with images. So what exactly do you mean "new grid in my image"? Do you want to modify the original image (bitmap)? If yes, how is this related to OCR?
If you're trying to modify the output document, what is the output format you saved the recognition result to?

After I get all OCR text results of a table,I create another new blank image and try to put all OCR text results and a new table grid in it.The new table grid has same layout and style as the source one.I just want to manage all elements (text/grid/line) in my new image,not just an 'IMAGE' .

Do you mean you want a function in our toolkit that returns the lines of the table grid?
If yes, you can use the AutoZoningCommand Class, which can return table data including locations of rows, columns and cells in the table.

Please check it and if it doesn't do what you want explain more about the requirements and what is different from this class.

I am finding a toolkit function or an alternative way to get a line's starting/ending Point object.Then I can draw a new table grid in a new image and fill it with OCR results to improve the poor source table grid quality.

Does this mean you solved the problem or not yet?
If not, please send me a sample image for which you're tying to locate the table grid lines and I will check if our toolkit can help you find these lines. If you send files, please put them in a ZIP or RAR file. You can post them here or send them to support@leadtools.com and mentions this forum post.

Here's a good quality image.

I will test with this image and give you the results of my test. Please send an email to support@leadtools.com to follow up.

I created this sample project to detect table cells in an image and then used DrawRectangle to draw over the table cells.
I just needed to use DeskewCommand to straighten your image.

I'll test it.Thank you.

Great Job.That's answer.
Thank you very much.

<BLOCKQUOTE><table width="85%"><tr><td class="txt4"><img src="/SupportPortal/CS/Themes/default/images/icon-quote.gif">&nbsp;<strong>Basel wrote:</strong></td></tr><tr><td class="quoteTable"><table width="100%"><tr><td width="100%" valign="top" class="txt4">I created this sample project to detect table cells in an image and then used DrawRectangle to draw over the table cells. <br>I just needed to use <span id="pagetitle">DeskewCommand </span>to straighten your image.<br></td></tr></table></td></tr></table></BLOCKQUOTE>

If you are using the OCR Professional engine and LEADTOOLS v17, then you can use the new IOcrTableZoneManager class. It allows you to detect the cells of a table in a document and return information such as cell bounding rectangle, its background color, border line style (including no border), border thickness and color.

LEADTOOLS v17 uses this information to re-construct the tables in the output document (such as PDF or DOC) as faithfully as possible.