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.
-
Check out this repository and download the source code
git clone [email protected]:silvery107/fast-photo-mosaic.git -
Install the required python modules
pip install -r requirements.txt -
Start photo mosaic by
import photo_mosaicand callingmosaic(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
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent