ICR and Bar Codes

ICR, which stands for intelligent character recognition, refers to methods used to understand handwriting. Considering that more than $10 billion is spent annually in the US for field coding documents, automating this process would seem particularly worthwhile. Many forms and applications are handwritten and automating the processing of these forms requires ICR. If ICR were working with sufficiently high rates of recognition then a document processing system could switch from manual field coding to fully- or semi- automated field coding. This form of business process automation generally pays for itself in very short order, e.g., less than six months.

Bar coding is the ability of a recognition system to automatically find and read an image bar code symbol. Bar codes are a very reliable way of transferring information within a computer system. For example, a robust way to port documents from one database to another is to generate a cover sheet for each document being ported where there is a bar code on each cover sheet than contains all the field information related to this file. Of course, this presumes that the original database has the values for each of these fields and it is simply a matter of transferring this information to the 2nd database.

CVISION's PdfCompressor v.4.0 release supports both ICR and bar code recognition. Of course, these two recognition technologies are not at all the same and, in fact, have very different recognition rates. On the one hand, bar codes are a very reliable, fault-tolerant method for transferring information in an image capture environment. However, this information, e.g., document field codes, needs to be available in electronic form at the time the document, including the cover sheet, is generated. So bar codes work very reliably and can be used as part of an integrated database solution or to automate a document port process.

ICR is used quite differently. In general, ICR recognition rates are not commercially viable. This is especially true if the information is in the form of unconstrained script handwriting. Recognition rates improve if the handwriting is constrained

i. Using boxes for each text character, or
ii. By being written in block characters.

Unlike bar codes, the information we need to ICR into the system has generally never been available in electronic form. So it is difficult to make ICR work in an automated or semi-automated environment. Often, using ICR processing as part of an automated or semi-automated solution requires an understanding of, and some engineering specifically for, the application domain. Certain fields might be strictly numeric or from a very finite set of values, e.g., car models. These constraints can be exploited with proper image engineering so that the recognition rates are viable in a semi-automated document processing workflow.

Both ICR and bar code technology are supported in PdfCompressor v.4.0. These recognition modules can be effective tools in automating your Company's document processing.

More Features: