Sustainability of Digital Formats: Planning for Library of Congress Collections

NOTE: This content was last reviewed and updated in 2004 and remains here for background information only. Beginning in 2015, the Library of Congress has published its format preferences as the Recommended Formats Statement, updated annually.

The illustrative tables presented here are intended to suggest how a curator of still images would determine format preferences. The first table illustrates a planning matrix that would result from analyzing the significant or essential characteristics for subcategories of still images. The second table illustrates how this analysis of significant features would be combined with technical information about formats to produce a set of format-preference statements for the various content subcategories.

Table 1: Significant characteristics of still image subcategories

1. For bitmapped images, clarity is typically represented by spatial resolution and tonal resolution (color depth). Spatial resolution relates to the number of pixels used in the two dimensions of an image; it is expressed either as absolute numbers or as ppi/dpi (pixels/dots per inch) in relation to dimensions on paper (original or intended). High-quality full-page magazine printing from a digital photograph requires images that range from 3,000x4,000 to 10,000x13,000 pixels.

2. The degree to which the color gamut represented in a given image can be managed and maintained through various outputs or migrations. This requires the use of formats with metadata that documents the color encoding used and provides additional data such as the color profile of the device used to create the image.  Additional elements may be needed in the case of multispectral images.

3. For still pictorial images, normal rendering includes onscreen viewing and rendering to print on a typical home or office printer. Normal rendering would also allow manipulation through software that allows image analysis, color correction, cropping, and scaling to permit fair use as "quotations." Normal rendering must not be limited to specific hardware models or devices and must be feasible for current users and future users and scholars.

4. To preserve digital content and provide service to users and designated communities now and decades hence, custodians must be able to replicate the content on new media, migrate and normalize it in the face of changing technology, and disseminate it to users at a resolution consistent with network bandwidth constraints.

5. Technical protection must not prevent custodians from taking appropriate steps to preserve the digital content and make it accessible to future generations. See Note 4.

6. Includes home photography and other images logically the same as I1, but with less artistic value for designated community.

7. Image files harvested from the Web through a program targeted specifically at capture of visual materials would be considered as I1, I2, or I3.

8. For images composed using vector graphics software, guidelines will be necessary (to be developed by custodians and user communities) as to when the functionality inherent in a vector representation is an essential characteristic and when the composition should be preserved as a bit-mapped graphic. Files in the proprietary native formats for commercial software used by professional graphics designers may include combinations of vector graphics and bit-maps, as well as a layering structure. Such files are expected to be reduced to bit-mapped images before being added to the Library of Congress collection.

Table 2: Format preferences for still image subcategories

1. Other formats exist and may be added as preferred or acceptable in the future, particularly if adopted widely. This document lists two ISO-related variants of the TIFF format (TIFF/IT and TIFF/EP), and a variant of PDF (PDF-X) designed to send "camera-ready art" to a printer. These formats are listed here in a provisional sense; the compilers are working to determine how widely adopted these formats may be.

2. The JPX level of the JPEG2000 standard supports more effective color management than the level 1 format (JP2). JPEG2000 offers many options for choices of quality level, and storage order for the encoded image data (codestream). Future investigation is needed to determine whether particular options should be encouraged or avoided when the objective is responsible long-term custody.

3. The degree of compression and specific encoding algorithms applied to images produce files at varying levels of quality. American Memory guidelines do not address this specific question but the project's experience with photographs suggests that JPEG compression ratios on the order of 5:1 provide very high quality when reformatting these continuous tone originals, while wavelet-based compression applied to maps suggests that this type of compression algorithm (also used in JPEG2000) can provide high quality, even when sharply defined lines or text fonts are present.

4. JPX_FF is preferred to JP2_FF because of its more explicit support for embedded metadata.

5. An example of a low-quality compressed image would be a photograph provided for the amateur by a photofinisher.

6. There is a general preference for sRGB color space over RGB color space when profiles or other color management tools have not been employed.