CBIR JPEG Study
Comparative study on Content-Based Image Retrieval from compressed images using different compression algorithms and content concepts. Compression: (jpeg, jpeg2000, jpegXR) Content: (SIFT, ...). To avoid artifacts from subsequent compression, an uncompressed dataset is used.
How it Works
You basically just have to run 'python doEverything.py 1 1 1'. Sit back and enjoy the show!
Requirements
You will need at least Python 3 with opencv and numpy/scipy.
Preferably, use the pip package manager to install these libraries.
As some of the algorithms (such as SIFT/SURF) are considered non-free, they have
been moved out of the base install of opencv and are only available in the
opencv-contrib package.
Luckily, there's package availabe to install with pip: opencv-contrib-python
Dataset
After you have installed the required software you will also need the following data sets:
Extract all datasets to analyze to a subfolder named data so that your direcotry structure resembles this: cbir/data/dataset1/class1/lotsoffiles.jpg.
Results
 |
 |
 |
 |
 |
 |
About the authors
This study was carried out by a group of students from the University of Salzburg under supervision of ...