Before beginning the scanning process consider a logical file naming system that the scanning supplier can apply to the images as they are created. You may want to browse and retrieve images by file name alone and a series of purely numbered images such as 0001 could make that very difficult. File names should be:

• Unique
• Consistently arranged
• Made up of letters and/or numbers
• The only non-alphanumeric characters used (if required) should be underscore _ to minimise issues with use across operating systems
• Not overly long or complex, to avoid character limitation issues in certain operating systems and also to avoid human error at the point of file naming

When choosing the correct format for digitised images, you should consider whether it allows lossless compression. This means the compression of the image without any loss of image data. This allows for a reduction in storage space required for your images whilst ensuring retention of all image data.

By comparison, lossy compression reduces the size of an image by discarding data which means data will be irretrievably lost i.e. even if you reversed the compression you would not retrieve all of the original image data.

JPEG2000 format is used by some institutions including The National Archives for a number of reasons including a greater level of lossless compression (than, for example, TIFF format) and because it is open source. However, to view JPEG2000 images you would need to download a free viewer such as IrfanView: https://www.irfanview.com/ as Windows does not come with a standard application that would enable you to open JPEG2000.

You may also want to consider baseline RGB TIFF 6.0 or higher format. TIFF is a popular image format that your scanning supplier should be comfortable generating. It has lossless capabilities and can be understood by a multitude of software.

If applying compression to your images to reduce file size, lossless compression is recommended.

• For JPEG2000 use lossless compression.
• For TIFF files we recommend ZIP compression, but LZW compression is also an option if preferred.

Lossy compression results in discarded information that can never be regained i.e. even if you reversed the compression you would not retrieve all of the original image data. It also
negatively impacts on the ability to carry out Optical Character Recognition (OCR), which enables the content of your scanned image(s) to be searchable.

DPI is dots per inch and is actually referring to the output resolution i.e. the dot intensity of the image when it is printed.

PPI is pixels per inch and refers to the input resolution of the image.

There can sometimes be confusion between PPI and DPI, particularly when manufacturers use these terms interchangeably.

The National Archives select PPI where possible. Use of lower PPI than required above can cause a loss of information, poor user experience and an inability for OCR to be successfully carried out on scanned images. PPI requirements vary according to the format of the material to be scanned and are detailed below.

• For ordinary documents, use a minimum of 300 PPI.
• For photographs, use a minimum of 600 PPI.
• For photographic transparencies, use a minimum of 4000 PPI.
• For microform aim for a resolution equivalent to a minimum of 300 PPI at the size of the original document. Take into account the reduction ratio usually recorded on the first frame of the microform).

Images should be single page, unless information crosses both pages. If a single scan cannot capture the page in its entirety, there should be sufficient overlap to allow users to determine clearly which of the separate digital images form the whole of the original paper page.

All scans should be size-for-size (for microfilm this refers to the size of the original), with a sufficient clear border/margin to demonstrate to users that the entire page has been captured.

For microform material, scan images in 8 bit grayscale using the Enumerated greyscale colourspace profile.

For all other material, scan images in 24 bit colour using the Enumerated sRGB colourspace profile.

All digital images should be legible and at least as readable as the original image from which they are derived.

All images should be viewed immediately after scanning as a check on satisfactory capture (for example images complete or not inverted) and rescanned if required.

While carrying out the scanning process the scanning supplier should be asked to capture a certain level of metadata regarding the scanning process. This ensures compliance with BS 10008 Evidential Weight and Legal Admissibility of Electronic Information – Specification which requires that, during the scanning process, ‘Associated scanning information (e.g. time, date, scanner number, number of pages) shall be created and retained’.

Additionally, capturing metadata related to the images created that will allow you to manage your records and ensure that there has been no loss or duplication during scanning.

As part of the metadata generation process it is best practice to document how you have generated and captured your metadata and what control procedures have been used to ensure its consistency and to guarantee authenticity.

The National Archives has created a tool which captures this type of information: DROID. It can be run over a batch of images to provide information on the files such as file name, file path, file format and extension. It will also generate a unique SHA 256 checksum for each file. A checksum is a ‘fingerprint’ for each file, made up of letters and numbers. No two checksums will be the same unless the content and format of a file are identical (the checksum does not relate to the filename, so if you had two differently named files with identical checksums, then you would know they were duplicate files in terms of content). You can generate new checksums for a file and check it against the original checksum, to ensure that the content of the file has not changed. A change in checksum will mean either someone has changed the content of the file and then saved it or that the file has been corrupted in some way. A change to the filename will not change the checksum. Checksums are not a good way to determine the intellectual content of images. Images of the same object will often have different checksums due to automated image capture information such as timestamp of image generation. It is important to generate a SHA 256 checksum after the image is created in order to be able to check that the file remains the same over time and to be able to determine if it has changed. This information assists both in managing the files and with digital continuity.

DROID cannot generate dates related to the original record from which the image has been created, so ensure you capture relevant dates relating to the record within the metadata as part of your metadata capture process. The National Archives provides further guidance on its website on digital continuity.

DROID is available to download for free from The National Archives website.

In order to achieve optimum results in OCR it is important that you follow the guidance on capturing at the appropriate PPI as listed above (or higher PPI than listed if preferred) and that if any compression takes place, it is lossless compression.

Some other points to focus on are:

• Use your high quality TIFF/JPEG2000 as the input for your OCR – rather than providing a compressed PDF.
• Ensure that your scans do not have noise being introduced by the scanning process – such as spots and dark patches which are not on the originals. These types of marks considerably reduce the quality of the OCR output.
• You must verify and assess the accuracy of the OCR output – do not assume it will be perfect because it is machine generated.
• Try to use OCR software which outputs to an industry standard XML format such as PAGE. This will provide a lot more metadata about the results, such as character coordinates, which could be useful in the future. It would also enable the use of software which can view the OCR output overlaid onto the original image which makes the verification of the OCR output process much easier.
• Ensure that the XML output is retained and can be connected to the image from which it was generated. This could be achieved by following the same naming convention for the OCR output as for the file to which it relates; but ensure you also add an extra identifier to the file name to show it is OCR so that you do not encounter issues with filename duplication with the image itself.