A Java wrapper around OpenCV with additions oriented toward medical imaging (window/level LUTs, segmentation primitives, raw image I/O). Packaged as an OSGi bundle that embeds the OpenCV Java classes and re-exports them, so consumers do not need a separate OpenCV dependency on the classpath. Used downstream by Weasis.

• Java 17 module (org.weasis.core.img) re-exporting org.opencv.core, imgcodecs, imgproc, img_hash, osgi, utils.
• PlanarImage / ImageCV — AutoCloseable abstraction over an OpenCV Mat with explicit release tracking.
• Image operations split into focused entry points: ImageAnalyzer (stats and measurements), ImageTransformer (geometric and visual operations), ImageIOHandler (read/write/thumbnails).
• DICOM-oriented LUTs and presentation state helpers (ByteLut, ColorLut, LutShape, LutParameters, WlParams, WlPresentation, DefaultWlPresentation, PresentationStateLut).
• Segmentation primitives (Region, Segment, ContourTopology, RegionAttributes).
• Raw image format with a 46-byte header (FileRawImage) and OpenCV-backed I/O.
• Native-library resolver (NativeLibrary) that maps os.name / os.arch to an OSGi-style <os>-<arch> spec and loads libopencv_java from an absolute path or by library name.

• JDK 17 or later
• Maven 3.6.3 or later
• One of the supported OS/architecture pairs for tests: linux-x86-64, linux-aarch64, macosx-x86-64, macosx-aarch64, windows-x86-64.

From the repository root:

mvn clean install

Build and test the bundle module only:

mvn -f weasis-core-img clean install

Run tests:

mvn -f weasis-core-img test

Run a single test class or method:

mvn -f weasis-core-img test -Dtest=ImageProcessorTest
mvn -f weasis-core-img test -Dtest=ImageProcessorTest#methodName

Produce a coverage report (JaCoCo, used by CI and SonarCloud):

mvn -Pcoverage -f weasis-core-img verify

Apply formatting before committing (google-java-format + EPL-2.0 / Apache-2.0 license header):

mvn -f weasis-core-img spotless:apply

Spotless is not bound to a build phase, so it only runs when invoked explicitly.

All OpenCV native dependencies are declared with provided scope. A downstream application is expected to ship the matching native library and load it at runtime by either:

• setting -Djava.library.path="path/of/native/lib" and calling NativeLibrary.loadLibraryFromLibraryName(), or
• calling NativeLibrary.loadLibraryFromAbsolutePath(path) with the absolute path to libopencv_java.

During tests, Maven unpacks the native library into weasis-core-img/target/lib/<os-name>-<cpu-name>/ and Surefire is configured to point -Djava.library.path there. The OS/arch profile in the root pom.xml selects which native classifier is unpacked.

Additional native binaries (other systems and architectures) are published in this Maven repository.

The artifact version tracks the underlying OpenCV release (currently 4.13.0), suffixed with a local revision (.1).

<dependency>
  <groupId>org.weasis.core</groupId>
  <artifactId>weasis-core-img</artifactId>
  <version>4.13.0.1</version>
</dependency>

A matching native classifier must be added separately, for example for Linux x86-64:

<dependency>
  <groupId>org.weasis.thirdparty.org.opencv</groupId>
  <artifactId>libopencv_java</artifactId>
  <version>4.13.0-dcm</version>
  <type>so</type>
  <classifier>linux-x86-64</classifier>
</dependency>

Other classifiers: linux-aarch64, macosx-x86-64 / macosx-aarch64 (type dylib), windows-x86-64 (artifactId opencv_java, type dll).

Two-module Maven build:

• Root pom.xml — BOM (weasis-core-img-bom) pinning weasis.opencv.version and declaring all supported native classifiers.
• weasis-core-img/ — the library itself (packaging bundle; the OSGI bundle plugin embeds dependencies inline under lib/).

Main source roots under weasis-core-img/src/main/java:

• org.weasis.core.util — generic helpers (file, string, math, stream, zip, Pair / Triple, SoftHashMap).
• org.weasis.opencv.natives — native-library loader.
• org.weasis.opencv.data — image data types (PlanarImage, ImageCV, FileRawImage, LookupTableCV, MetadataParser, ImageSize).
• org.weasis.opencv.op — image operations (ImageAnalyzer, ImageTransformer, ImageIOHandler, ImageConversion, ImageContentHash). The legacy ImageProcessor facade is kept for backward compatibility but is deprecated — new code should use the split classes above.
• org.weasis.opencv.op.lut — window/level and presentation LUTs.
• org.weasis.opencv.seg — segmentation primitives.

The OpenCV Java sources themselves are not vendored: they are pulled at build time from the opencv:sources artifact, unpacked into target/sources-import, and added as an extra source root via build-helper-maven-plugin.

• Formatter: google-java-format via Spotless.
• License header (EPL-2.0 OR Apache-2.0) is enforced by Spotless on every Java file.
• Tests use JUnit 6 (org.junit.jupiter.*) and Mockito.

Dual-licensed under the Eclipse Public License v2.0 and the Apache License v2.0.