Building document scanner with opencv and python

Author: Ravindra Sadaphule

Using some interesting combinations of techniques such as edge detection , finding contour and perspective transform, we can build a decent scanner application for document. It is useful for scanning pictures of receipts or pages in book as shown below. Later I’ll also show you example of how we can make this scalable by performing bulk scanning.

Here are basic steps in creating a scanner application in python

Step 1 -> Edge Detection : Edge detection is finding points where there are drastic changes in intensity. These points gives us edges of objects on the screen.

Code:

Step 2 -> Finding Contour : We then use edged picture from above to draw the contour around receipt image with green border as shown below. This contour is crucial to understand orientation and 4 important points (top left (tl), top right(top right), bottom right (br), and bottom left (bl)

Here is the code:

Step 3: Perspective Transform : Once we establish tl, tr, bt and bl all we need to do is to rotate and scale these points to get birds eye view of the umage by using perspective transform.

Here is the code