Image-processing / machine learning - Removed coffee stains (noise) from dirty documents

Developed multiple image processing and machine learning approaches in R to remove noise in the scanned documents and ensemble using Xgboost to improve the output quality of the OCR process of digitizing documents.

Tools - R Techniques -

1. Edge detection and morphology (BiOps package)
2. Median filtering
3. Adaptive thresholding
4. Feature engineering - Identifying gaps between sentences
5. XgBoost ensemble

"ImageCleaning.R" - R code - impelemted algorithms for image processing. "XGboost.R" - Complements "ImageCleaning.R" ensembling different techniques. "TeamsmaRt_Project1_Image_Processing.pptx" - White-Paper for the project. "TeamsmaRt_Project1_Image_Processing.pptx" - Powerpoint Presentation of project. "XGBoost_Workspace_Image.RData" - RData file with preinitialized variables (to save time to load the images in vectors) "train_sample" - Sample images folder.