Introduction to Image Processing
jongoesboom
Posts: 26 ✭✭✭
Introduction to Image Processing
“Photoshopping” has become one of those ubiquitous terms we’ve come to know to mean – manipulating a digital image for one reason or another. In the case of Grooper, it’s called Image Processing and it’s used to accomplish a similar goal. This cleanup is especially applicable to scanned images. With scanned images, there can be any number of issues we may want to clean up. From de-skewing and border cleanup to adjusting brightness and contrast, Grooper’s Image Processing capabilities are vast. Another amazing feature of Grooper’s Image Processing is its ability to detect features like stamps, barcodes and shapes. The recognition of these features is important because it can tag a page with specific information (or metadata) and manipulate it in a specific way later in the process.
Image Processing can be done in Grooper by creating an IP Profile and applying it to an Image Processing step, Scan step, or OCR Profile. When creating an IP Profile, you have access to Grooper’s entire library of 50+ Image Processing steps. With all of these options available, IP Profiles can easily get complex quite quickly. However, with Grooper’s interface you can see step by step what is happening to the image and easily rearrange, delete, or add steps to accommodate your needs.
Permanent vs. Temporary Cleanup
There are a few methods to consider when using Image Processing. One method is to run Image Processing to make permanent changes to the source document. This method is primarily used to clean up artifacts such as speckles, border clean up, auto-orienting, or cropping the images that were originally introduced during the scanning process. Permanent Image Processing is an archival way to clean up documents so that they can be viewed as closely to the original document as possible.
Another method is a temporary cleanup. This method is used to get a document cleaned up to go through a Full Text OCR step. This usually involves steps such as line removal, removing blobs, thresholding, removing halftones, or even dropping out a certain color to make the image easier for the OCR engine to process. By attaching the IP Profile to the Full Text OCR step, it will run a temporary cleanup of the image before running OCR and then revert back to the image that the Full Text OCR step started with.
A third method of using Image Processing is shape detection. The human mind recognizes shapes quickly and easily by replacing detail with a simple general profile of the object.
Grooper shape detection works on this same principle. Filtering is used to produce simple block outlines of a reference picture allowing detection of the object on low resolution images. After being trained with a reference image, Grooper can detect its location and orientation in an image. This is great for finding stamps, seals, logos, postmarks, etc. The saved stamp information can then be used for classification or extraction purposes.
