Bengalore (Karnataka) [India], July 26: Extrieve has announced an updated release of the ImageWizHelper SDK which simplifies the imaging operations for developers. The new black and white conversion feature help to remove the shadows from images and results in more accurate OCR data extraction.
Tags/keywords: Shadow removal, OCR Capture, Image preprocessing.
Feature details
There are several threshold algorithms that can be used to convert grayscale images to black and white; which is the image depth used for OCR. Global threshold, Niblack threshold, Otsu threshold, and Sauvola threshold are some of the different standard algorithms available. Standard imaging libraries support most of these algorithms. Based on the type of input images and the context of the usage, the developer has to choose the right algorithm.
For some images with very dark shadows or light exposures, additional preprocessing like contrast increase and sharpening may be required to improve the edge detection.
With an increase in the use of mobile devices to capture images, these challenges are encountered very often in applications.
The OCR Black & White conversion feature in the ImageWizHelper SDK performs the necessary preprocessing to improve the text quality of the image.
This feature has been tested and optimized with images captured from a wide
variety of mobile captured devices. This can result in up to a 20% improvement
in the OCR recognition rate.
Sample Image with shadows (redaction done to protect data confidentiality)
After black and white conversion using Otsu algorithm
After using ImageWizHelper OCR BW conversion.
ImageWizHelper
ImageWizHelper provides a simple API to compress the images from any type of source and standardize them to specific layout sizes. It compresses images to a smaller size while maintaining the readability of the documents. Multiple images can be combined and converted into a single PDF or Tiff file. For uncompressed files, SDK can reduce the size up to 90% and for compressed it can reduce the size up to
50%. QuickCapture is a similar product for mobile capture and compression use cases.