Processing of Large Image Data in Fiji: ImgLib2/BigDataViewer Ecosystem

ImgLib2 is a Java library for processing multidimensional data. It is a main infrastructural component of Fiji, KNIME, and Icy, which together serve the majority of image processing needs in life sciences. The library provides low-overhead and extensible abstractions over data type, dimensionality, and storage. For example, in Fiji, ImgLib2 provides unified access to “normal” in-memory images, arbitrarily large transparently cached images, as well as legacy data structures. ImgLib2 is widely used and its API has remained stable for several years. This project will maintain and further develop the ImgLib2 library, and support and grow the surrounding developer and scientific user community. Work on the library will therefore focus mostly on consolidation, and improve Python interoperability and support for transparent (partial, on-demand) image replication over the network.

BigDataViewer (BDV) leverages ImgLib2 to provide efficient visualization and processing of the largest image datasets assembled to date. BigDataViewer is a re-slicing viewer supporting multi-modal time-series with multiple views, tiles, channels, multi-resolution pyramids, etc. It supports transparent caching, on-the-fly generation, and interactive deformation of images. BDV is used in numerous Fiji plug-ins for visualizing and processing big multi-modal image data sets, generated by for example light-sheet or electron microscopy. Although the learning curve remains steep, the ImgLib2/BDV projects have a strong developer community with deep knowledge of big image data processing. This project will reinforce this ecosystem by funding core development, lowering entry barriers to attract new developers, and supporting user outreach, as biologists are beginning to realize that their bioimages are beyond the capabilities of existing software solutions, yet they may not be aware of the big data solutions offered by ImgLib2/BDV in Fiji.