Simon has been involved in software development since the days of paper tape. He has developed niche software for information management.
Nearly every flatbed scanner comes with sophisticated control software for cropping, removing dust and scratches, de-skewing images and many other functions. However, many flatbed scanners (particularly multi-function devices that also serve as printers) treat scans of multiple photos as a single image. Such images may come from scanning a photo album page, or from scanning multiple photos placed directly on the scanner.
A common task is to turn each of the multiple photos into a separate digital image, and optionally to add a description and edit the image itself. Although this can be done manually with supplied scanner software or easily available image editors, such as Windows Photos or IrfanView, it’s laborious.
The latest release of the Epson Perfection software (Epson Scan 2) has built-in functionality for automatically creating separate images from separate photos placed on the scanner in Photo Mode but as of May 2021 is only available on Macs.
Automating this process requires that the application detects the boundaries of photos contained in an image, with minimal or no parameter adjustment. This is a difficult image segmentation problem, as photo backgrounds may not be easily distinguishable from the photo content.
With the exception of ImageMagick and Caption Pro, applications do not give any indication of how they achieve this, but any algorithm will fail on some images, and a manual option for defining the boundaries of photos is desirable in this case. It is difficult for applications for devices using touchscreens to achieve this level of interaction with images, so they rely on automatic segmentation, which works successfully for the majority of cases.
Either a scanner or a camera can be used for creating digital images from existing physical images. The two methods and many other issues are discussed in detail on Ken Watson's excellent website All About Digital Photos.
1. Using a Scanner
Small A4 flatbed scanners are readily available, and may even be incorporated into multi-function devices mainly used for printing. Individual hard-copy prints can be placed on the scanner bed and the scanner directly accessed from image splitting software.
Good results are usually obtained in this way as the background to the multiple photos is highly uniform and easily distinguished from the photo content, provided the photos are well-separated.
If the photos are on a photo album page, the process is more difficult as photo albums commonly have a patterned background, and the pages may be yellowed around the edges, making the background non-uniform and the extraction process more difficult. A reflective transparent layer to protect photographs from handling may also be present.
Mounting corners for photos may also be used, making the photo content non-rectangular. The page size may also exceed A4 dimensions.
2. Using a Camera
Taking a photograph of a number of individual photos is much easier than using a scanner, and may be the only option if the photos are within a glass-covered frame with a matte, such as shown below or a photo album with large pages.
The extraction process is more difficult for already grouped photos, as the background may be patterned or non-uniform, and the oval mattes often used give make individual photos non-rectangular.
Where the photos are behind glass, reflections may degrade individual image quality. These can be minimized by illuminating the grouped photos brightly and making the camera as dark as possible using oblique, directional illumination from at least two light sources.
3. Using Smartphone Camera Apps
The availability of high-resolution smartphone cameras has encouraged the development of apps for processing images of paper photo prints taken with the device. These apps generally offer auto-cropping, and de-skewing of a photo placed on a uniform background, and may offer glare removal and perspective correction.
Individual photos can be saved as individual files, but the resolution of these will be limited by the resolution of the smartphone camera and the number of photos included in a single image. An A4 flatbed scanner scanning at 300 dots per inch creates a 29-megapixel image—more than twice as many pixels as images from a 12-megapixel camera (such as is found on the iPhone XS and Google Pixel 3) and is clearly preferred for high-resolution scans.
Consider Pixel Dimensions
Pixel dimensions for smartphone cameras are not necessarily an indicator of image quality. Pixel dimensions may be much larger than that required by the image produced by the optical system, as storage is cheap and pixel dimensions are a good marketing feature. In one case, 48-megapixel images from a smartphone could be resampled to 8 megapixels with no loss of image quality, with a 6-fold reduction in image size.
Apps may also offer easy grouping and sharing of extracted photos via social media and the web.
Most smartphone apps suggest restricting the number of photos included in an image, as the number of pixels used in each extracted image will be somewhat less than the camera pixel count divided by the number of images.
Large Batches of Small Images
For images containing large numbers of small photos (such as old photo albums), pixelated extracted images may result from processing an image of the whole page. In this case, sections of the page or even single images should be used. If only single photos are included in the image many photo scanning apps are available, such as Google’s PhotoScan (Android and iOS) and Best Cool Apps’ Photo Scanner (iOS only). These generally offer the same facilities as the multiple photo apps reviewed here.
A high-resolution digital camera may also be used to create files which are then processed by a desktop application. As most desktop applications are built for use with flatbed scanners, they will not have the perspective correction.
All applications provide guidance on the scanning process, the major point being to keep individual photos well separated.
A Summary of the Applications Reviewed
1. ImageSplit, Caption Pro (Windows) 4.5 stars
ImageSplit provides access via a scanner as well as from files, optional manual splitting controls, good performance with individual photos. Well-suited to large projects. Caption Pro (which includes ImageSplit) provides excellent captioning facilities and includes ImageSplit.
2. Caption Pro for Mac (Mac) 4 stars
Similar capability to Caption Pro/ImageSplit but without face recognition. Includes image splitting capability.
3. RansenScan (Windows) 3.5 Stars
Optional manual control of splitting, direct scanner access, single-line captions. Good for individual photos placed on the scanner but not for camera images or scanned album pages. Slow operation. Simple interface but best suited to tech-savvy users.
4. ScanSpeeder (Windows) 4.5 Stars
Direct access to flatbed scanners, optional manual control, highly robust splitting algorithm, and simple interface. Works well with photographs of album pages as well as scanned images.
5. AutoSplitter (Windows) 4 Stars
Optional manual control of the splitting process, direct access to flatbed scanners, robust splitting algorithm, and control of splitting parameters. Clunky interface. Captioning facilities are very basic.
Like Autosplitter, but without any manual control or scanner access. Extraction algorithm not robust. Good metadata setting in SnipTag.
7. PhotoMyne ( iPhone, iPad, Android)—5 Stars
Simple, slick, powerful, and expensive. Provides many of the facilities of Google Photos.
Similar products are suitable for well-conditioned images. PicScanner Gold offers a range of extra functionality.
9. PhotoShop, Photoshop Elements (Mac, Windows Desktop)—2 Stars
Great for image editing and metadata set, but slow and bulky to install. Bundled products may not be desired. Poor extraction performance and no parameter control.
10. GIMP with Divide Scanned Images Script (All platforms)—2 Stars
Free, multi-platform but poor performance. Only suitable for software enthusiasts.
11. GIMPScan (Mac only)— 3.5 Stars
Reasonable performance but installation and GIMP interface daunting. The best option for Mac Users prepared to deal with GIMP.
12. VueScan (Windows, Mac)—2 Stars
Very basic: no captioning, skew correction, or manual frame definition, but may be able to access scanners that other applications cannot.
If you are placing individual photos on a scanner, Caption Pro (for Windows or Mac) is accurate for automatic splitting, has excellent captioning facilities, and is well-suited to large projects. RansenScan gives is also very accurate, but processing may be slow. ScanSpeeder detects photos well for uniform backgrounds, but may not perform de-skewing accurately. Autosplitter may not detect all individual photos.
If you are scanning album pages or using camera images containing multiple photos, different products have widely differing performances on different images. If you are keen to optimise processing on images of this type, different products should be evaluated. RansenScan appears to require a non-patterned uniform background to individual photos and is not suitable for album pages or camera images.
Smartphones and tablets offer convenience, but extracted individual images from apps may lack resolution when used with common devices. Photomyne has outstanding performance but its automatic image enhancement may not always be desired. PicScanner or PicScanner Gold may be adequate at a lower price, especially if only a few photos appear in images.
The Products in Detail
Note that the scanner interfaces of desktop applications were not tested, due to the wide range of scanners available.
Caption Pro (v 3.1.119), ImageSplit (v 1.0.14), Windows
Caption Pro, from Aleka Consulting, is a newcomer to multi-photo extraction, which is provided by the ImageSplit application. It is similar to ScanSpeeder, Autosplitter, and RansenScan in its automatic photo detection and split definition capability but it has superior captioning capabilities and can generate captions using face recognition. It also has a subdivide option for detected splits, making it well suited to processing contact sheets or negative strips containing multiple images of interest.
It can add two-part captions which can be edited after application, and it can use images from the backs of photos as captions. Caption Pro comes with a free 30-day, 25-caption, or splits save limit demo license. A license for permanent operation and 1 year of major updates costs US$29. It runs on 64 bit Windows 7,8.1 and 10, but face recognition is only available on Windows 10.
Although the product is called Caption Pro, the photo extraction is performed by a separate application called ImageSplit, which is bundled with Caption Pro and accessed through it. It is also available as a standalone application for use where captioning is not required. A standalone ImageSplit license costs US$19. It comes with a free demo license allowing up to 25 splits to be saved over a 30 day period. Input can come from a scanner or from an image file. ImageSplit provides a choice of automatic detection algorithms ( as well as manual split definition) and the option to pad a loaded image in order to improve performance where the photos touch the edge of the scanned region.
Unlike most other dedicated applications, the documentation describes the algorithm for separating the photos from the background and indicates its limitations and how they can be minimized. The screen after automatic photo detection is shown below.
Automatically detected photos (or splits) are outlined in red and can be adjusted for size, position and rotation and splits can be added manually if photos are not detected. Splits can be rotated, inverted, sharpened and have their contrast and brightness adjusted individually or collectively. The position of the background pixel (which should be outside any photo) can be adjusted by dragging a circle, and the sensitivity of the detection can be adjusted in the settings screen.
After the definition of the splits, the splits can be saved. Captions can then be added, optionally using the name of the split file created. The separation of the captioning from the splitting operation allows bulk captioning of any number of splits obtained from multiple scans. Captioning is highly flexible, providing unlimited length and the ability to edit captions after application. Images can be used as captions, and logos added if required.
Captions can be created from detected faces to which names have been added which are recognised in images. There are two options for face recognition. Standard mode processes all data locally and is included in the basic license.
Premium mode uses a commercial web service for detection and recognition and requires an Internet connection. It offers improved speed and accuracy over Standard mode. 50 Premium face recognition transactions are included in the Basic license, with additional Premium transactions costing between US$0.02 and US$0.03 per transaction.
The aspect ratio of the splits can also be adjusted to ensure that all of the splits are included in hard-copy printouts.
Context-sensitive help is provided via a local Windows help file. The help includes detailed advice on getting the best results from scanners and cameras.
ImageSplit is well suited to automatic processing of multiple photo images with uniform or slowly varying background, using a scanner or a camera. Its choice of splitting algorithms allows it to work with photos on a patterned or non-uniform background. The separation of splitting and captioning features makes Caption Pro/ImageSplit particularly suitable for large-scale projects requiring the digitization of hundreds of paper prints and the face recognition feature is useful for large collections of family photos.
My rating: 4.5 stars
Caption Pro (v 3.1.17) Mac
The Mac version of Caption Pro from Aleka Consulting has most of the functionality of Caption Pro (Windows) but lacks face recognition, but parameters of the photo extraction process are editable and the process gives accurate results. It is available through the Apple App store with a free 30-day, 25 caption demo license. A full license for permanent operation and 1 year of major updates costs US$29.
Splitting can be performed using the Mac scanner interface software or the Apple Computer Vision rectangle detection software. Detailed help suggests methods of improving performance with black-and-white images, which are often poorly handled by multiple photo processing software. The scanner interface requires MacOS 10.13 (High Sierra) or later, but other features will work on MacOS 10.9 (Mavericks) or later.
My rating: 4 stars
ScanSpeeder (v 2.19) Windows
ScanSpeeder comes from the Canadian company NitroLogic and has a very similar interface to Autosplitter; which product came first is not obvious. The start screen is shown below.
Pro and Standard versions cost US$39.95 and US$29.95 respectively for permanent licenses. Both versions include updates for 1 year. The Pro version includes support for 48-bit colour and 16-bit black and white scans.
ScanSpeeder can operate directly from a scanner, or from files containing images of multiple photos. After loading a file, extracted photo boundaries are marked in green and the extraction can be performed by clicking Extract Photos, to give the results shown on the right-hand side of the screen as shown below.
The extraction algorithm is very robust, working successfully on images with a non-uniform or patterned background using auto-background detection as shown below. However, it stumbles when the background is a smooth image gradient and de-skewing may not be accurate if individual photos are placed on a scanner.
Manual adjustment of background is also possible by selecting a background region via the gear icon beside the Detect button, but this was not necessary even for the range of backgrounds shown above.
The Help text for adjustment of the automatic extraction parameters does not refer to controls visible in the current version. Selecting a different background colour using the dropper function did not appear to work.
Photo boundaries can be adjusted by clicking within the frame. Sides can then be dragged, or the entire frame rotated or deleted. New frames can be added via the Add Frame button. Captions can be added to extracted photos by adding in the [Add optional text/tag] area. Text is saved in the Windows Title and Subject field when the tick icon is clicked.
Caption size and font and colour can be set in the Settings area accessed via the gear icon to the right of the Extract Photos button. The area used by the caption is increased with caption length and font size, but multiple line captions are not supported—all caption text is displayed continuously. Unicode captions (as used for Asian languages) and emojis can be shown, although emojis lose their colour.
Captions are toggled via the T option visible in the extracted photo:
An Enhance button increases brightness and contrast for selected extracted images. Selected Images can also be rotated by an arbitrary angle using the Rotate button. Selected images can also 90, 180 or 270 degrees via the menus in the image shown above.
Extracted files can be saved in either JPEG or TIFF format
Somewhat terse help is available via a local file rather than via a URL, making ScanSpeeder easily usable without an Internet connection, and there are a number of online tutorial videos, articles and blogs available.
ScanSpeeder is an excellent product, with a highly robust automatic extraction algorithm allowing it to work with photographs of old photo album pages as well as scanned images of multiple photos. Its captioning facilities are good and a simple interface makes it suitable for non-technical users.
My Rating: 4.5 stars
RansenScan (v 4.1.4—Windows)
RansenScan comes from Italy-based company Ransen Software, which offers a number of CAD and Graphics programs for Windows. It costs US$29.99 for a permanent license. The demo version adds a watermark on any extracted photos but is otherwise fully functional.
As the installer for the demo version does not use a code-signing certificate, a warning message appears during installation on Windows 10. The website may also be incorrectly flagged by anti-virus software as containing adware.
The RansenScan startup screen has a combination of a conventional text menu structure and a row of buttons for common functions as shown below. Tips are optionally shown on startup.
RansensScan can import directly from a scanner or from already scanned files. The File menu options to select a scanner and to initiate the scan are separate, which may cause confusion.
The same device may also appear multiple times. Loaded images are rotated as specified by the EXIF orientation flag, if present. Images are automatically split into sub-images, which are straightened and displayed on the right of the screen as shown below.
Automatic recognition options allow specification of the minimum and maximum size of photos and a maximum aspect ratio.
Clicking on a sub-image highlights the region automatically selected, which can then be adjusted if required:
If no sub-images are detected automatically, selection regions can be defined manually by clicking and dragging over the displayed image as shown below.
Clicking the ‘A’ Button or selecting the 5th button from the left creates a sub-image from the defined area:
The process of selecting an area and creating a sub-image can be repeated as often as required.
Sub-images can be rotated or have frames and captions added by in various positions by right-clicking and selecting from a context menu, as shown below. Caption text is shown as a single line, truncated if the text at the selected caption size exceeds the photo width:
Colour adjustments to the image can also be made via the context menu.
Although RansenScan automatic extraction is intended to work on images from scanners, it will work with uniform background images from cameras (as found on mobile phones) but the process may be very slow.
Extracted sub-images can be saved via the File menu in a number of common image formats. Bulk processing of multiple images and watermarking of extracted images are also supported. Comprehensive help is available via a web link.
RansenScan works well for processing scans of individual photos, with a fairly simple interface. The automatic extraction is not robust enough to deal with scanned album pages. However, the ability to manually define photo boundaries makes it usable where automatic extraction fails. The addition of frames makes for more pleasing results, but captioning facilities are limited.
My Rating: 3.5 stars
AutoSplitter from Hungarian company Chimera Creative Studio is a sophisticated product for processing scans of multiple photos. Permanent registration costs US$19.99 for 1 year of upgrades or US$29.99 for 2 years of upgrades.
The unregistered version has full functionality but all image displays are watermarked. The start screen is divided into an area showing the loaded image, and the extracted photos (or splits):