Receipt recognition algorithm

Internship project with OpenCV

The algorithm is able to improve shadow/lights on image, crop the receipt itself, fix skew and detect text via Tesseract engine.

Steps:

• Find and crop image with help of morphological operators
• Denoising and illumination normalization
• Converting to grayscale, gamma, brightness contrast correction and increasing sharpness
• Converting to black and white with adaptive threshold
• Deskewing
• Select text areas and Tesseract them

Source code is located in target/, where client is http client, server is http server and image_core is recognition algorithm itself. Samples located in sample_images.
Also you can find images of each step in misc/process (not ordered!).