Skip to content

Latest commit

 

History

History
85 lines (63 loc) · 3.79 KB

DIPlib.md

File metadata and controls

85 lines (63 loc) · 3.79 KB
Raw
layout title author
default
DIPlib
Cris Luengo

DIPlib

DIPlib is an extensive C++ library for quantitative image analysis. It has hundreds of algorithms for manipulating, filtering, segemeting and quantifying images. There currently are more than 3000 documented symbols.

There are other image processing/analysis libraries available, some of them hugely popular. Why do we keep investing time in developing and improving DIPlib? The short answer is that we believe DIPlib offers things that are not available elsewhere. The library is built on the following three principles:

  1. Precision

    We implement the most precise known methods, and output often defaults to floating-point samples. The purpose of these algorithms is quantification, not approximation.

  2. Ease of use

    We use modern C++ features to provide a simple and intuitive interface to algorithms, with expressive syntax, default values, and little boiler-plate code required from the user. There is no need to be aware of an image's data type to use the algorithms effectively.

    Furthermore, developing an image analysis program involves a lot of trial-and-error, rapid prototyping approaches are applicable: the edit-compile-run loop should be quick. We aim for short compile times with pre-compiled algorithms and few public templates.

  3. Efficiency

    We implement the most efficient known algorithms, as long as they don't compromise precision. Ease-of-use features might also incur a slight overhead in execution times. The library can be used in high-throughput quantitative analysis pipelines, but is not designed for real-time video processing.

Algorithms in DIPlib typically accept input images of any data type (though, of course, some algorithms are specific to binary images, or cannot handle complex images, etc.) and any number of dimensions (algorithms that are limited to one specific dimensionality typically show so in their name). The image data type and dimensionality do not need to be known at compile time. Images can have pixels that are vectors or matrices, for some examples on how this relates to the three points above, see "Why tensors?".

There are many other unique things about DIPlib, we encourage you to explore the documentation to learn more about it.

See also the examples/ directory for a series of simple C++ programs that demonstrate how to use various features of the library.

Modules

Besides MATLAB and Python bindings, DIPlib 3 currently has interfaces to the following packages:

  • Bio-Formats: DIPjavaio is an interface to Java-based image readers. It is designed to allow DIPlib to read hundreds of image file formats through OME Bio-Formats, but is generic enough to be used with other Java libraries as well.

  • OpenCV: a single header file provides copyless conversion to and from OpenCV images, for OpenCV version 2 and newer.

  • Vigra: a single header file provides copyless conversion to and from Vigra images.

The DIPlib project also contains DIPviewer, an interactive image display utility.

Installation

We recommend that you build the library yourself from sources. There are some optional external dependencies, all required dependencies are included in the repository. For build instructions see Building the DIPlib library.

When linking against the DIPlib library without using CMake, it is important to set a few preprocessor macros. See Linking against the DIPlib library for details.