The Key Technologies Driving Digitisation: From Bit Depth to OCR

Digitisation is more than placing a document on a scanner and pressing a button. Behind every clean, readable scan of an old manuscript or a fragile newspaper lies a set of technical decisions that determine whether the digital copy is faithful, usable, and worth preserving. Three technologies sit at the heart of this process: how finely an image captures detail and colour, how files are compressed to save space, and how scanned text is turned into something a computer can actually read. Understanding these technologies helps library and information professionals make smart choices that balance quality, cost, and long-term access.

Bit depth and resolution: the foundation of image quality 🔗

When we talk about the quality of a digital image, two technical attributes do most of the heavy lifting: resolution and bit depth. They are often confused, but they measure very different things. Resolution describes how much detail an image captures, while bit depth describes how much colour or tonal information each part of that image holds. Both directly affect quality, and both increase the file size as they go up.

What resolution really measures 🔗

Resolution refers to the number of points of information a scanner or camera records about an image. The more points captured, the more detail the image holds. It is usually expressed in pixel dimensions, such as 4000 by 5000, or in dpi and ppi. These two acronyms are frequently used interchangeably, but they are not the same. Dots per inch (dpi) measures the number of ink dots a printer produces, while pixels per inch (ppi) measures the number of pixels in an image within a one-inch space. For digitisation work, ppi is the more accurate term, even though “dpi” remains common in everyday conversation.

A higher resolution captures finer detail but also increases file size. For documents and photographs, a resolution of around 300 dpi or more is considered a common baseline for good quality, while archival projects often use higher settings to preserve every detail. It is worth remembering that resolution alone does not guarantee a high-quality image. Standards bodies treat resolution as one specification among several, including colour space, file format, and bit depth, that together define the characteristics of a digital file rather than its quality on their own.

How bit depth shapes colour and tone 🔗

Bit depth, also called colour depth or sample depth, refers to the number of bits used to describe the colour of a single pixel. The higher the bit depth, the more colours or shades the image can represent. The relationship is exponential. A 1-bit image gives each pixel only two possible values: black or white. An 8-bit grayscale image produces 256 shades of grey, which is enough that the human eye cannot detect any tone changes between them. A 24-bit colour image, which uses 8 bits for each of the red, green, and blue channels, can reproduce around 16.78 million colours.

This is why different materials call for different bit depths. The choice depends on the content being captured rather than a desire to always use the highest possible setting:

1-bit (bitonal): Best suited to high-contrast documents such as clean printed type, where each pixel only needs to be black or white.

8-bit grayscale: Recommended for printed text, manuscripts, hand- or typewritten papers, and photographs where colour is not important.

24-bit colour: Recommended for archival-quality colour images, capturing a spectrum beyond what the human eye can perceive.

Higher bit depth improves colour reproduction but, like resolution, increases the data size of the file. A common practice in preservation work is to keep a high bit-depth master file, often 16-bit per channel, while saving smaller 8-bit copies for everyday access. As heritage imaging specialists note, the main downside of higher bit depth in a derivative file is simply that it makes the file larger, so institutions reserve it for the preservation copy.

Calculating file size 🔗

Resolution and bit depth together drive how much storage a digital collection requires, which matters enormously for planning. The file size of an image can be estimated by multiplying the image width, image height, the square of the resolution, and the bit depth, then dividing by eight to convert bits to bytes. A single 24-bit colour page scanned at 600 ppi can take up roughly 100 MB, which means a collection of just over 22,000 pages can demand more than two terabytes of storage. This calculation is exactly why the next technology, compression, becomes so important.

Compression: balancing quality against storage 🔗

As collections grow, the cost of storing uncompressed files becomes a serious consideration. In the world of libraries, archives, and museums, conversations now routinely involve terabytes and even petabytes of data. Compression is the technology that reduces file size, and it comes in two fundamental types: lossless and lossy. The difference between them is one of the most important concepts in digitisation.

Lossless compression 🔗

Lossless compression reduces file size by finding patterns and abbreviating the underlying binary code, but it does not discard any of the original information. When the file is opened, every pixel is restored exactly as it was. This makes lossless compression the preferred choice when exact data preservation is critical, which is why it is widely used in digital libraries and medical imaging.

The TIFF (Tagged Image File Format) is the workhorse of preservation. It can be stored uncompressed or with lossless compression, and it is considered the industry standard for high-quality archival storage. PNG is another lossless format, valued for preserving logos, line drawings, and screenshots where sharpness matters. The trade-off with lossless compression is that the files remain large compared to lossy alternatives, even after compression.

Lossy compression 🔗

Lossy compression takes a different approach. It permanently removes some image data to achieve a much smaller file. The technique works by discarding information that is least likely to be noticed by the human eye, a field sometimes described as psycho-visual modelling. The JPEG format epitomises lossy compression. It uses a method called the Discrete Cosine Transform to throw away data the eye is unlikely to miss, which is why JPEG works beautifully for photographs but poorly for text or logos, where lost data shows up as blurriness.

The most important rule with lossy compression is that the loss is permanent. Once data is discarded, re-saving the file at a higher quality setting will not bring it back. For this reason, compression should never be applied as an automatic default. It should be a considered choice made fresh for each project, weighing whether maximum quality or maximum space saving matters more.

When lossy compression can actually help 🔗

It might seem that archives should always avoid lossy compression, but the reality is more nuanced. The British Library, when digitising its newspaper collection, adopted the JPEG 2000 format with lossy compression and gained two benefits. It dramatically reduced file size compared to TIFF, and it also reduced image noise, which in turn improved optical character recognition results. However, further research showed that the same level of compression could not safely be used for material with a high colour depth. This illustrates the central lesson of compression: there is no single correct setting, only choices that suit specific materials and goals.

OCR: turning images into searchable text 🔗

Capturing a high-quality image of a printed page is only half the job. A scanned image is, to a computer, just a picture of text. It cannot be searched, copied, or edited. Optical Character Recognition (OCR) is the technology that bridges this gap, converting scanned images of text into machine-readable, searchable digital documents.

How OCR works 🔗

OCR analyses the patterns of light and dark that make up characters on a page and uses algorithms to identify the patterns that form letters, numbers, and symbols. It then compares these patterns against a database of known characters and assigns a value to each one. The software also uses context clues such as word spacing, punctuation, and grammar rules to improve accuracy.

The process usually begins by cleaning up the scan, which can involve correcting skew, reducing noise, and identifying different segments of the page. The OCR engine then reads the text using one of two main methods. Pattern recognition compares characters against a stored database of patterns and works best on highly standardised documents. Feature extraction uses a more sophisticated approach, breaking each character down into its components of lines, curves, and intersections, which makes it better at handling varied fonts.

Why OCR matters for libraries 🔗

For library and information work, OCR is what transforms a folder of scanned images into a genuine digital library. The technology embeds a layer of recognised text on top of, or alongside, the original image, which means the original appearance is preserved while the content becomes fully searchable for reference and research. A user can search for a specific word and instantly find every page that contains it, an ability that paired with good indexing and metadata makes a collection vastly more useful.

The benefits extend further. OCR enables the creation of searchable digital archives, reduces the need for physical storage, supports remote access, and improves accessibility for visually impaired users through screen readers. It is worth noting that OCR is not perfect. Even advanced systems can produce errors that require proofreading, especially with poor-quality originals or unusual fonts. This is precisely why the earlier technologies matter so much: a clean, well-resolved scan produces far better OCR results than a noisy, low-quality one.

How these technologies work together 🔗

These three technologies are not isolated choices but parts of a single connected workflow. The resolution and bit depth you select determine how much detail and colour you capture, which sets the ceiling on quality. The compression you choose decides how that quality is balanced against storage cost and whether your master files remain faithful. And the quality of those captured images directly shapes how accurately OCR can read the text. A poorly scanned page compressed too aggressively will frustrate OCR, while a well-captured, lightly compressed image will read cleanly. Good digitisation is the art of making these decisions consciously, with the long-term goal of preservation and access always in mind.

What do you think? If you were planning to digitise a fragile, century-old handwritten manuscript versus a stack of modern printed reports, how would your choices about bit depth, compression, and OCR differ for each? And where do you think the balance should fall between preserving perfect quality and managing the very real costs of storage?

References

1. https://www.adoc-solutions.com/en/post/the-best-image-quality-in-digitisation-projects
2. https://records-express.blogs.archives.gov/2023/08/22/when-we-say-we-want-resolution-dpi-and-ppi-explained/
3. https://www.goodreads.com/author_blog_posts/23606374-dynamic-range-and-resolution-for-digitization-projects
4. https://heritage-digitaltransitions.com/in-depth-on-bit-depth/
5. https://academiccommons.columbia.edu/doi/10.7916/D83R126F/download
6. https://museumsandheritage.com/advisor/insights/digital-imaging-is-jp2-the-file-format-of-choice-for-the-museum-and-heritag/
7. https://www.sciencedirect.com/topics/computer-science/lossless-image-compression
8. https://www.papercut.com/blog/print_basics/what-is-ocr-optical-character-recognition-and-how-it-works/
9. https://andersonarchival.com/learn/what-is-optical-character-recognition-ocr/
10. https://digi-texx.com/techblog/what-is-optical-character-recognition/