libmpeg7

Library implementation of MPEG-7 digital image descriptors calculation and comparison algorithms. This project was created as part of an engineer's degree thesis. As of now, after many years I'm trying to make it somewhat working again. As this was originally a project to incorporate old MPEG7 library in new code, there are many parts (especially algorithms) that were directly copied as-is from that library. I might or might not refactor it with new implementation.

About MPEG-7

MPEG-7 is a multimedia content description standard that provides a rich set of standardized tools to describe multimedia content. This library focuses on the visual descriptors specified in the standard, which can be used for content-based image retrieval, image classification, and similarity-based image search.

Supported Descriptors

The library implements the following MPEG-7 visual descriptors:

Color Descriptors

• Dominant Color - Provides a compact description of the representative colors in an image or region
• Scalable Color - A color histogram encoded using the Haar transform for scalable representation
• Color Layout - Captures the spatial distribution of colors in an image
• Color Structure - Identifies localized color distributions using a structuring element
• CT Browsing - Describes color temperature information in an image

Texture Descriptors

• Homogeneous Texture - Characterizes texture regions in an image using frequency domain statistics
• Texture Browsing - Browsing type descriptor for texture
• Edge Histogram - Represents the spatial distribution of edges in an image

Shape Descriptors

• Region Shape - Describes shapes using Angular Radial Transform (ART)
• Contour Shape - Describes shape contours using curvature scale-space representation

Each descriptor can be configured with specific parameters to adjust its behavior. The descriptor IDs can be used with the command line application and API calls to specify which descriptor to extract.

Building the Library

Prerequisites

• C++14 compatible compiler
• CMake (minimum version 3.5)
• Java Development Kit (JDK) for JNI support

Build Instructions

The project includes a build script that makes compilation straightforward:

# Clone the repository
git clone https://github.com/yourusername/libmpeg7.git
cd libmpeg7

# Make the build script executable
chmod +x build.sh

# Build both debug and release versions
./build.sh -all

The build script supports the following commands:

• ./build.sh -all - Builds both debug and release versions
• ./build.sh -dbg - Builds only the debug version
• ./build.sh -rel - Builds only the release version
• ./build.sh -c - Cleans the project (removes build directories)
• ./build.sh -h - Shows the help message

After a successful build, the libraries and executables can be found in:

• Debug version: build/x64/Debug/ or build/x32/Debug/ (depending on your architecture)
• Release version: build/x64/Release/ or build/x32/Release/

Usage

Command Line Application

The library comes with a command-line application (mpeg7_app) that allows you to extract descriptors from images and calculate distances between descriptors.

Extracting Descriptors

To extract a descriptor from an image:

./build/x64/Release/mpeg7_app extract <descriptor_type> <image_path> [param_name param_value ...]

Where:

• <descriptor_type> is a number from 1 to 10 representing the descriptor type:
  • 1: Dominant Color
  • 2: Scalable Color
  • 3: Color Layout
  • 4: Color Structure
  • 5: CT Browsing
  • 6: Homogeneous Texture
  • 7: Texture Browsing
  • 8: Edge Histogram
  • 9: Region Shape
  • 10: Contour Shape
• <image_path> is the path to the image file
• Optional parameter name-value pairs can be provided depending on the descriptor type

Example:

./build/x64/Release/mpeg7_app extract 3 image.jpg NumberOfYCoeff 64 NumberOfCCoeff 64

Calculating Distances

To calculate the distance between two descriptors stored in XML files:

./build/x64/Release/mpeg7_app distance <xml_file1> <xml_file2> [param_name param_value ...]

Where:

• <xml_file1> and <xml_file2> are paths to XML files containing MPEG-7 descriptors
• Optional parameter name-value pairs can be provided depending on the descriptor type

Example:

./build/x64/Release/mpeg7_app distance descriptor1.xml descriptor2.xml

This command compares the two descriptors and outputs the distance (similarity measure) between them. Lower distance values indicate greater similarity between the descriptors.

Library Integration

C++ Integration

The library can be integrated into your projects either as a static or shared library. Include the main header file in your code:

#include "Mpeg7.h"

Java Integration (JNI)

The library includes JNI bindings that allow it to be used from Java applications. The following methods are exposed:

// Extract descriptor from an image file
public native String extractDescriptor(int descriptorType, String imagePath, String[] parameters);

// Extract descriptor directly from image data
public native String extractDescriptorFromData(int descriptorType, byte[] imageData, String[] parameters);

// Calculate distance between two descriptors
public native String calculateDistance(String descriptorXml1, String descriptorXml2, String[] parameters);

To use the JNI interface:

1. Ensure the shared library (.so on Linux, .dll on Windows) is in your library path
2. Load the library in your Java application using System.loadLibrary("mpeg7")
3. Create an instance of the class that implements these native methods
4. Call the methods with appropriate parameters

Reporting Issues

If you encounter any errors or issues while using the library, please create an issue in this GitHub repository

Author

Krzysztof Kucharski