Integrating OpenCV for Native Image Processing in Phimpme Android

OpenCV is an open source computer vision library written in C and C++. It has many pre-built image processing functions for all kinds of purposes. The filters and image enhancing functions which we implemented in Phimpme are not fully optimized. So we decided to integrate the Android SDK of OpenCV in the Phimpme Android application.

For integrating OpenCV in the application, the files inside the OpenCV Android SDK have to be properly placed inside the project. The build and make files of the project have to be configured correspondingly.

Here, in this post, I will mention how we fully integrated OpenCV Android SDK and configured the Phimpme project.

Setting Up

First, we downloaded the OpenCV-Android-SDK zip from the official release page here.

Now we extracted the OpenCV-Android-SDK.zip file to navigated to SDK folder to see that there are three folders inside it. etc, Java and native.

All the build files are present inside the native directory and are necessarily copied to our project. The java directory inside the SDK folder is an external gradle module which has to imported and added as a dependency to our project.

So we copied all the files from jni directory of SDK folder inside OpenCV-Android-SDK to jni directory of the Phimpme project inside main. Similarly copied the 3rdparty directory into the main folder of the Phimpme project. Libs folder should also be copied into the main folder of the Phimpme but it has to be renamed to jniLibs. The project structure can be viewed in the below image.

Adding Module to Project

Now we headed to Android Studio and right clicked on the app in the project view, to add a new module. In that window select import Gradle project and browse to the java folder inside the SDK folder of OpenCV-Android-SDK as shown in figures.

Now we opened the build.gradle file of app level (not project level) and added the following line as a dependency.

  compile project(':openCVLibrary320')

Now open the build.gradle inside the openCVLibrary320 external module and change the platform tools version and  SDK versions.

Configuring Native Build Files

We removed all files related to cmake be we use ndk-build to compile native code. After removing the cmake files, added following lines in the application.mk file

APP_OPTIM := release
APP_ABI := all
APP_STL := gnustl_static
APP_CPPFLAGS := -frtti -fexceptions
APP_PLATFORM := android-25

Now we further added the following lines to the top of Android.mk file

OPENCV_INSTALL_MODULES:=on
OPENCV_CAMERA_MODULES:=on
OPENCV_LIB_TYPE := STATIC
include $(LOCAL_PATH)/OpenCV.mk

The third line here is a flag indicating that the library has to be static and there is no need for external OpenCVLoader application.

Finally, find this line

OPENCV_LIBS_DIR:=$(OPENCV_THIS_DIR)/../libs/$(OPENCV_TARGET_ARCH_ABI)

in OpenCV.mk file and replace it with

OPENCV_LIBS_DIR:=$(OPENCV_THIS_DIR)/../jniLibs/$(OPENCV_TARGET_ARCH_ABI)

And finally, add the following lines to java part. So that the OpenCV is initialized and linked to application

static {
   if (!OpenCVLoader.initDebug()) {
       Log.e( TAG + " - Error", "Unable to load OpenCV");
   } else {
       System.loadLibrary("modulename_present_in_Android.mk");
   }
}

With this, the integration of OpenCV and the configuration of the project is completed. Remove the build directories i.e. .build and .externalNativeBuild directories inside app folder for a fresh build so that you don’t face any wrong build error.

Resources

Continue ReadingIntegrating OpenCV for Native Image Processing in Phimpme Android