Less is More

Anyone processing images automatically in batch mode needs lean, easy-to-use tools. The program of choice is usually convert from the ImageMagick suite [1]; however, this program has some drawbacks. In particular, convert does not necessarily impress with its stability. This is not typically a major issue, but in case of complex command lines, you might see the program calls abort with a (mostly incomprehensible) error message and produce incorrect results, or even none at all.

In fact, even composing a command line for convert proves to be a task that should not be underestimated: The program supports no fewer than 300 different options that often exhibit strange syntax. Many options serve very specific tasks and require experience and a great deal of background knowledge on the part of the user – for example, do you know offhand what calling convert … -resize 200\>x300^ … would give you?

Many users also complain about the lack of effectiveness of the ImageMagick tools. The programs are not especially quick, nor do they deliver particularly brilliant results – in part because of the massive functional diversity, which is often not needed.

The GraphicsMagick [2] suite is an attempt to resolve these issues; another competitor goes by the name of ExactImage [3]. Both provide only a subset of ImageMagick's repertoire and restrict their feature set to the most important functions.

What differences exist between the three suites? Although ImageMagick comes with almost 20 different programs, GraphicsMagick is a single program – along with three configuration tools; ExactImage provides seven tools (Table 1).

Although GraphicsMagick tries to cover most of ImageMagick's functions, the tiny ExactImage suite delivers only one universal tool (econvert, as a counterpart to convert), as well as some programs for very specific tasks, such as reading barcodes and detecting blank pages.

And, Action!

Like convert from the ImageMagick tools, econvert is the main tool when it comes to creating new images from existing ones. The syntax follows the structure:

$ econvert -i <input file> -<option> ... -o <output file>

The main thing here is to pay attention to the correct order of the individual components; otherwise, econvert (just like convert) might ignore the specified options. Starting with a single input file, you can use multiple switches -<option> -o <output file> to create as many output files as you like on a single command line.

The way the ExactImage tools work is completely controlled by various options, just like ImageMagick (Table 2). ExactImage supports only a subset of the available image formats, and they differ in terms of input and output (Table 3). That said, ExactImage does support the main image formats to a large extent.

When scaling with --scale, the algorithm automatically determines which method to use. If you are not satisfied with the results it delivers, you can set the method yourself. The options in this case are:

• --bicubic-scale
• --bilinear-scale
• --box-scale
• --ddt-scale
• --nearest-scale

Moreover, the --thumbnail fast filter generates thumbnails.

The example in Listing 1 once again illustrates the use of econvert. It produces three different output files from one input file.

Who Are You?

The second important program from the ExactImage suite is edentify, which is used to identify the format of an image. The example shown in Listing 2 is deliberately difficult: Identifying raw data formats is already a challenging task, and this is again aggravated by the DNG created by the conversion. Listing 3 shows the way ImageMagick handles this job with identify.

Although edentify provides reasonably correct information and does not even try to guess file sizes, it outputs many error messages. ImageMagick's identify takes a different approach. First, it converts the raw data to a bitmap format, which it then analyzes.

Whether this approach makes sense is doubtful. The file sizes specified by identify are totally wrong because they relate to the bitmap files generated by the conversion, rather than the originals. Evaluating a DNG as a TIFF can be seen as reasonably correct, because DNG was designed as an extension of the TIFF standard.

The ability to display specific information about images with edentify is interesting. This is achieved by specifying a "format string" as an argument for the -f option. For example, -f "%f" gives you the file name, -f "%w" the width, and so on. Table 4 summarizes the most important variants. To connect several strings, you can use \t (tabulator), \n (newline) or \r (carriage return).

All told, edentify offers only limited features; again, you might need to resort to ImageMagick's identify.

Utilities

The other programs from the ExactImage suite are more specialized. For example, bardecode automatically evaluates scanned bar codes; this still works even with relatively poor images. The e2mtiff tool merges multiple images into a multipage TIFF file:

$ e2mtiff x1.jpg x2.png ... -o <out>.tiff

Multipage TIFFs are not particularly widespread nor particularly popular on Linux. However, Gimp can open and edit them. Figure 1 shows what a multipage TIFF looks like in Gimp when you load it. You can now individually edit the pages, which exist as layers (Figure 2), or convert them to new images by dragging them over the toolbox with the mouse.

Figure 1: On opening multi-page TIFFs, Gimp asks you which pages to open as layers.

Figure 2: You will find the pages as normal layers in the layers dock.

The optimize2bw program prepares scanned pages to improve their suitability for subsequent processing with OCR. The empty-page program detects blank pages, so you can discard them up front. Finally, hocr2pdf is the probably the most-used tool in the ExactImage program suite.