Fast Photo Mosaic

In this work, I implemented a photo mosaic algorithm based on feature matching. I designed a feature descriptor based on the mean histogram of the LAB color space, applied the K-D tree to match the color blocks of the target image sub-region, and used the pre-computed feature pool to optimize the synthesis speed, and realized the mosaic photo that has better performance than Foto-Mosaik-Edda.

Quick Start

1. Check out this repository and download the source code

   git clone git@github.com:silvery107/fast-photo-mosaic.git

2. Install the required python modules

   pip install -r requirements.txt

3. Start photo mosaic by import photo_mosaic and calling mosaic(tgt_img_pth, tiles, types)

Parameter	Description
tgt_img_pth	Directory path of target image.
tiles	The resolution of mosaic elements, and each values should be an integer multiply of 8.
types	Currently it support two types, natural and manmade, you can add new image types under data/<your_type> with corresponding <your_type>.txt description file.

• For more usage details, please check quick_start.py

Composited Image Gallery

Call Graph