Image Size Limits

This topic summarizes the changes to image size limits over the last versions.

Version 12 imposed both global and module-specific limits. The global limits for loading images were 8400 pixels and 22 inches for both width and height. Some of the OCR modules had lower hard-coded limits: 5440 pixels for MOR and 6688 for MTX, with MTX being also limited to an image area to 20 megapixels - practically around 4000 x 5000 pixels.

Version 15 saw fundamental changes in the image handling algorithms of all modules to allow them to deal with practically unlimited image sizes: 32000 pixels in both dimensions. This limit was set because image coordinates must fit into a signed short integer. In version 15 however, there was still a code segment that checked the size of images being loaded and returned an error if the images exceeded certain limits either in pixel dimensions or in actual length and/or height (pixel dimensions divided by DPI). These limits were determined by settings that could be changed by the user (see Kernel.Img.Max.xxx in the section List of all the RecAPI settings in the RecAPI help, under 'Related pages').

Kernel.Img.Max.Pix.X 8400

Kernel.Img.Max.Pix.Y 8400

Kernel.Img.Max.Inch.X 28

Kernel.Img.Max.Inch.Y 28

The goal of these default values for the image size limit was to prevent the time consuming processing of huge images; for example, when the kit was used in an unattended background process recognizing all the files on a machine. If an application was required to process images bigger than these above limits, they could be reset to any value (up to 32000 for the pixel count). See section Settings Manager Module in the RecAPI help.

Version 16 retains the limit of 32000 pixels and the default of 8400 pixels, however, the limitation in inches is removed, to avoid the rejection of large but low resolution images, whose pixel count is still acceptable. In exporting large pages, be aware that Microsoft Word does not accept RTF files with either lateral larger than 22 inches.

It is important to understand that memory limitations on a given system may limit the size of the images that can be successfully processed and that this may restrict image size to below the 32000 pixel width and height limits.

The following table summarizes testing done with version 15 when recognizing pages using the PLUS3W module in Accurate mode and exporting to XML formatted output. European page sizes were used - for each move from A4 to A3 to A2 and so on the page area doubles. The results for A4 practically apply for the US page size 'Letter' (8.5 x 11 inches).

These values are indicative only, since performance is strongly influenced by characteristics of individual image files, the processor used and the computing environment. These tests were done on a hyper-threaded machine with 2 x 2.4 GHz and 1GB of RAM.