In 1997 Wayne Rasband, computer scientist at the National Institutes of Health (NIH, Bethesda, USA), released the first version of ImageJ. It is the successor of the image processing software NIH Image, which was well known at this time. NIH Image was exclusively tied to the Macintosh platform, the new software, however, is independent from platforms since it is completely programmed in Java, the lingua franca in the computer world. Since it is open source and maintained on the NIH server [1], everybody may freely download a bundled standard version and start working within a few minutes.

Almost every week a new version of ImageJ is released with improvements and fixed bugs. The integrated online update function of the software makes it easy to keep it always up-to-date. Meanwhile it has gradually evolved to an amazingly powerful tool for almost any kind of image processing task though it has been kept well structured and flexible. Moreover, a countless number of plugins and macros have been made available by a vast community of scientific users who help to solve simple or even very sophisticated problems of image processing.

At the beginning, the application of ImageJ was almost entirely restricted to microscopy. Today medical imaging like computer tomography, astronomy, cartography, and interpretation of aerial images have also gained strong importance. An overview over ImageJ is given here.

Working with ImageJ

ImageJ exists in different versions, the standard version provided by the NIH server or for example Fiji [2], which represents a standard version of the ImageJ plus a certain collection of plugins for 3D processing and image registration.

When ImageJ is launched first by a user, a surprisingly small application window is displayed (see fig. 2). It is hard to imagine that this is housing so many powerful image processing tools. The working logic of the software is based on opening as many image files as desired. The number of open images is only limited by the computer memory which was initially reserved for ImageJ.

Image processing tools are then applied to the image window which actually has focus. This may be a single image or a slice of an image stack.

Image Formats

Image files may consist of single images or image stacks. Both may also contain meta data. Almost any type of image is supported from 8 bit to 32 bit binary or even 3x16 bit RGB. The normal ImageJ version opens and writes image files in standard image formats such as TIFF, JPG, PNG, BMP, FITS and some more. When adding the Bio-Formats plugin [3] about 115 more mostly proprietary file formats can be read, among them almost all formats once introduced by known microscopy and scientific instrument companies. In medical imaging the DICOM format plays a prominent role. So the great DICOM plugin to view and also write to DICOM files has to be mentioned here [4].

Some Image Processing Tools Which Are Giving Power to ImageJ

a) Image Calibration
Images can be calibrated for example in physical distance units and optical units like Optical Density. Different curve fitting methods are offered for the latter.

b) Measurements
Geometrical parameters like length, area of regular and irregular shapes, perimeter and angle can be measured and are presented in tabulated form. This is simply performed by drawing the appropriate selection contour in a non-destructive manner into the image.

More complex measurements are possible when a classical blob analysis is carried out. Interesting structures are first isolated from the background by applying a threshold. Structures appearing above threshold are then automatically measured and also counted with respect to a number of parameters (i.e. mean gray, integrated density, Feret‘s diameter, kurtosis, skewness, area fraction and much more) which were selected before. This method is extremely useful in particle analysis but can also be used for fast cell counting.

Moreover special tools are available for the analysis of electrophoresis gels and blots.

c) Filtering, Transforms, Segmentation and Background Subtraction
A number of filters are present as in any normal image processing software for sharpening, blurring etc. Own transform matrices can be defined and applied. Fast Fourier Transform (FFT) is also found in the option list. Moreover, there is a broad variety of segmentation algorithms acting on images converted to 32 bit binary. For example, these can perform the identification of edges, skeletonizing, dilating, and eroding or sophisticated segmentations like Watershed and Voronoi. The rolling ball model with some selectable options is implemented for background subtraction.

d) Image Calculator
The image calculator is composing a result image from 2 source images by applying an operator such as add, substract, divide, XOR or some other. For example, in the field of fluorescence microscopy quite often 2 originally black and white images are taken from a double-labeled specimen with different optical filters. Each image is then colorized by adding the appropriate Look-Up-Table (LUT; ­ImageJ comes with numerous LUT already defined). These 2 Images can then be merged by the action of the ADD operator to give a single image showing both labels and the overlapping areas in a different color. This complex appearing procedure is done with a few mouse clicks or even more conveniently by a small macro script (see below).