KONICA MINOLTA

About Konica Minolta

Giving Shape to Ideas

Image compression (compact PDF)

The digital multifunction machines for business become have various functions to enhance the value of each model. Among those the functions, the technology to convert scanned document image into PDF file which is smaller than existing JPEG compressed image has been focused with the increasing cases of using colored documents. With this technology, communication time of "Scan to mail" sending scanned image by e-mail can be reduced. As the result, it can be easier to use the digital multifunction machines.

Generally speaking, when file compression is performed, the quality of image becomes worse. However, in our development project, we have succeeded in the technology to convert into small file size of PDF with keeping the image quality by separating text and photo from the scanned image, and compressing each of them separately.

Conversion process of Compact PDF is as follows;.

*
To separate text and photo accurately from the document image.
*
To perform an optimize compression for each text and photo.
*
To convert each text and photo into each layer and put them together. And to unite them as a PDF file format.
With this process, it is possible to convert into the smaller file size than conventional simple compression.


Fig.1 Concept of Compact PDF

The features of compact PDF conversion technology are as follows;

  1. Automatic separation of a document image into "Text" and "Photo" areas (area separation by composite detection of text and photo)
  2. The text area and photo are each applied to the suitable compression separately, and are converted into PDF (compression and conversion into PDF)

Area separation technique by composite identification of image and text

As for the separation of photo and text in the document, many different methods have been researched and developed. The area separation of the photo and text is the most important technology in the compact PDF conversion technique, and this performance greatly depends on this technique. We have developed this separation algorithm based on our own great research experiences in image processing field.

The most important feature is to detect photo and text with each criterion, to analyze the result and to separate them more accurately based on the correlation. So far, an existing separation is not performed accurately because of detecting and separating them based on just one criterion.

In the photo area, even small photo area sometimes can be separated accurately by analyzing various features. So far, small photo images could not be distinguished from text. However, we have succeeded in improving accurate detection of photowith this technology.

In the text area, only text can be accurately separated using the text layout information based on the technique of text extraction technique which is our experience of photocopier development. As the result, this technology has greatly improved the following errors;

*
Detect the part of image as text mistakenly.
*
Impossible to detect text liked to image.
*
Detection is affected by back ground color of paper.


Fig.2 Photo/text separation

Even if accurate photo/text separation is performed, text composed of multiple colors used to become wrong color. Our technology can extract unnatural portion by the color and the size of each letter from the area recognized as text and modify accurate coloring of letters. Fig. 3 shows a sample. In the conventional processing illustrated in Fig. 3-(2), "W" and "r" are linked with other letters, and they are identified as a single letter, and all the letters become black.

In our technology, each letter is separated to be correct color, as shown in Fig. 3-(3).


Fig. 3 Resultant letters

Compression and conversion to PDF

After separation of photo and text, each part is modified into the most suitable format to the PDF layer configuration. The text is binarized and is performed to G4 compression used in a fax machine. The photo is modified to have an adequate resolution and is performed to JPEG compression. Each of them as a layer is converted into a PDF file.

In the "Scan to Mail" sending a scanned image by e-mail, our new technology allows you to send compact and high-quality images when transmitting a plurality of colored documents which are not used often because of size problems. Furthermore, because of no impact to OCR, it is expected to provide security tools related to archive or copy log.

Today, we are researching and improving quality and performance of technology to separate photo and text which are our core technology with a global method for layout detection.

Back to top