Dr. Tom's Independent Software Reviews

Image Localization:
This chapter will be quite short as I do not want to repeat the excellent article in MLC&T #58: "Image Localization and New Technology" authored by Ultan Ó Broin. While all web enabled image file types (GIF, JPG and PNG) are flat or single-layered (the same goes for print media types like TIF and PICT), they are in most cases developed using multi-layered image authoring systems such as Adobe PhotoShop, Jasc Paint Shop Pro or Ulead PhotoImpact. In these applications, text can be either represented as a vector-based text layer or transferred into a bitmap to achieve special effects. Whereas there is a certein chance to extract text layers, there is no way to get to text transferred into artwork. But even accessing text layers is limited to the PSD format of Adobe Photoshop, which can also be created by most of the other professional image editors. There is, however, only one tool currently available to do this trick: Rainbow, within the RWS Localization Tools, developed by the RWS Group.

Fig. 1: Image Authoring Using Multiple Layers in Adobe PhotoShop:
Accessible text information is stored in vector based text strings as in Layers 2, 3 and 4.
Please click on the image for a closer look!

Rainbow etracts text layers from the PSD format and prepares them for translation by wrapping the extracted strings separately either in the pure XLIFF format or into XLIFF with special RTF layers suitable for translation within the translation memories TRADOS workbench or Wordfast. Unfortunately, as in all proprietary file formats, the developer of Rainbow was not told exactly how a PSD file is build, therefore the Photoshop filter that Rainbow uses is a very basic one. It only knows where the text layer starts and where it ends. Because of this, when merging back the translated file, the new text has to be shorter or equal in length to the original, otherwise the PSD file would be corrupted. In our example in Fig.1 & 2, you can also see another problem of this approach: Different color coding of words is not adopted to the translation but left at the same byte position - therefore blue encoded "meinen" in Fig. 1 becomes blue "my revi" in the result (popup window in legend of Fig.2).
Waiting for a 'real' Photoshop filter, Rainbow also accelerates and streamlines the traditional "copy+paste" method by createing a LKP file if one or more string could not be merged. You can use this lookup file with Album to automate more the process of copying and pasting.

Fig. 2: Rainbow created XLIFF with three text strings from the Photoshop file in Fig. 1
Please click on the image for a closer look!
Click to see the result of saving back the translation into the PhotoShop file!

Even though it is not capable to extract text strings from image files, the Globalization Image Assistant or short GIA can be a helpful tool to facilitate the traditional Copy & Paste method and store and organize huge amounts of multilingual images from large localization projects. As in the case of XLIFF, GIA stores all informations including the inserted translatable strings into a XML file format. This would be a very nice option to batch translate many image files but unfortunately, all text from one image is merged into one single string which limits its use. -Relative locations = broken links. +Statistics. +media browser, +individual image comments and instructions, +qa compares tool image size etc MAKE it SHORT!!

.

Fig. 3: Media Bowser and Project Statistics of the Globalization Image Assistant
Please click on the image for a closer look!

Many professionals and experts have strong expectations towards the SVG-format to end most of the problems encountered in image localization. As the Scalable Vector Format is another application of the XML file format and stores text strings as plain text, every XML-enabled translation software can painlessly be used to localize SVG graphics. There are, however, a few aspects which still temper too much enthusiasm about this format: First, complex bitmapped content is still imported in JPEG or other pixel-based format, which will prevent any text elements within these components to be extracted. Secondly, as other promising technologies such as VRML and SMIL (see below - this article), after years of development, the use of SVG is still quite uncommon and only a few specialized authoring packages such as Jasc WebDraw offer access to the rich opportunities of SVG such as basic animations. As Adobe, Corel and lately also Microsoft show a strong support of SVG, the acceptance of this great format will hopefully become more wide spread soon.

Fig. 4: SVG Authoring using Jasc WebDraw
Please click on the image for a closer look!
If your browser supports the display of SVG, click to view the example!