Delphi Developers Tools - Tiff images and image compression

TIFF images and image compression

by Kent Phelps

Background
TIFF stands for Tagged Image File Format, a specification for storing computer raster images in such a fashion that the information about how the image is constructed (and thus, how to decode it) is stored in a series of TAGs, or small record structures, within the image itself. Tags describe such things as image width, height, compression method if any, the author or copyright, origin of the image, etc. TIFF was originated by companies working together such as Aldus, Hewlett-Packard, Silicon Graphics, Xerox, and so forth.

These days people generally expect TIFF images to be groupIV fax black and white documents. While that is a common use, especially for document imaging systems, there is much more involved than that.

Compression
It is important to separate the idea of the TIFF format for an image in general from the compression scheme of the image. A TIFF G4 image is one which is compressed using the CCITT group IV bi-level compression technique, and wrapped in a TIFF image whose tags explain this to a TIFF reader. The image could just as well be stored without the TIFF format and tags, but the reader would need to know ahead of time how wide and deep an image is, and what colors are in the image before decompressing it. Many "proprietary" image systems do just that. They use a standard compression type, but a nonstandard header to describe it, so normal TIFF readers (software) cannot show the image.

Typical compression types are CCITT (international fax standard) Group 3, either 1 dimensional (1D) or two dimensional (2D), group 4 (group IV), Huffman , RLE (Run Length Encoded), LZW (Lemple-Ziv-Welch - owned by Unisys Corporation).

For details of exactly what a TIFF image should be, get a copy of the TIFF specification from Aldus Developer's Desk.

Imaging issues
For document imaging, noteworthy issues are the compression scheme, dots per inch (resolution), photometric interpretation (black/white), contrast, multi-page formats, and intelligent use of tags.

Compression scheme. These are almost always GroupIV. This seems to be the best, and smallest image, for black and white business documents. Group 3 is not bad and is the format used by many fax machines, but there are several flavors of group 3 and not all readers can read all group 3 images, even though the tags tell how to do it. This issue normally arises in conjunction with use of tags for exporting an image so that other systems can read the images. Some document systems keep images in a internal format but allow export in a common format like GroupIV, while others use group IV as a native format
Dots per inch. The resolution used for business documents generally is 200 dpi. Laser printers normally print at 300 dpi. Printing a document scanned at 200 dpi on a 300dpi printer does not improve the inherent resolution, yet often the goal in an imaging system is to strike a good balance between storage space and image quality. For some uses, 150 dpi would be acceptable where the documents are not detailed, and for other uses where there is a lot of fine print 300dpi is necessary. Going from 200 dpi to 300 dpi does not increase the image size by 50% (1.5 times) but by 2.25 times (1.5 x 1.5) since the image expands in both dimensions. Some fax documents (standard fax) are created as 200 x 100 resolution and are expanded vertically during display. This can also save storage space if the documents are suitable. Color images are big no matter what and increase the complexity of the compression technique. JPEG (Joint Photographic Experts Group) compression is effective, but it is a "lossy" compression. With JPEG, either some detail is lost while colors are preserved in reducing image size, or else the detail is kept but the image is larger. If the image is stored initially at a certain detail level, for the most part it cannot be improved later.
Photometric Interpretation. This refers to how the image is displayed. Are the 1's black and the 0's white, or the other way around? Generally, this is not something the end user has to worry about unless the imaging system will be exporting images to another product. In that case, they both need to agree on black and white.
Contrast. It is often possible while scanning documents, especially poor ones and handwritten ones, to improve the readability of the resulting image using scanner features. Some scanners such as Fujitsu have electronic circuits which dynamically adjust contrast as the documents are scanned, evening out variations, and can sometimes even be used to emphasize or de-emphasize certain parts of the image like shiny ball-point pen markings to create a better image than the original.
Multi-page formats. TIFF (along with GIF and DCX formats) includes the ability to "package" several images into one file. Usually in document imaging each page is a separate image file, but when exporting images, especially if sending a set of images over the internet or in e-mail, it is sometimes helpful to be able to group all the images together so they cannot get lost. Many readers do not understand how to decompress any but the first image, however.
Intelligent use of tags. The tags in a TIFF image normally just store information for the software to read to display the image, but it is possible to put other things in tags, such as indexing information. This means that it is never possible to lose the index to an image even if the databases controlling it are lost. Timestamps and information about where and how the image was created can also be permanently placed in the image. This is not a crucial feature, but nice to have sometimes.