Abstract
Images and scanned text documents are gradually
more used in a vast range of applications. To reduce the needed
storage or to accelerate their move through the computers
networks, the document images have to be compressed.
Traditional compression mechanisms, which are generally
developed with a particular image type and purpose, are facing
many challenges with mixed documents. This paper describes a
statistical block-based technique for an automatic document
image segmentation and compression. Based on the number of
detected colors in each region of the image, this approach creates
a new representation of the image that can produce very
highly-compressed document files that nonetheless retain
excellent image quality. The proposed algorithm segments the
compound document image into blocks of equal size. The blocks
are classified into seven different categories. Each category
represents an image part that shares the same properties. A new
representation of each category is formed and the similar adjacent
blocks are merged to form labeled regions sharing the same
properties. At the end, to achieve better compression ratio, the
different regions of the image are compressed using different
compression techniques.
more used in a vast range of applications. To reduce the needed
storage or to accelerate their move through the computers
networks, the document images have to be compressed.
Traditional compression mechanisms, which are generally
developed with a particular image type and purpose, are facing
many challenges with mixed documents. This paper describes a
statistical block-based technique for an automatic document
image segmentation and compression. Based on the number of
detected colors in each region of the image, this approach creates
a new representation of the image that can produce very
highly-compressed document files that nonetheless retain
excellent image quality. The proposed algorithm segments the
compound document image into blocks of equal size. The blocks
are classified into seven different categories. Each category
represents an image part that shares the same properties. A new
representation of each category is formed and the similar adjacent
blocks are merged to form labeled regions sharing the same
properties. At the end, to achieve better compression ratio, the
different regions of the image are compressed using different
compression techniques.
| Original language | English |
|---|---|
| Number of pages | 9 |
| Journal | International Journal of Engineering and Advanced Technology |
| Volume | 6 |
| Issue number | 4 |
| Publication status | Published - 30 Apr 2017 |