The goal of this project is to implement processing pipeline for detecting logos in provided image. In this project I detect Microsft logo. All necessery algorithms are implemented from scratch.
- Load image to memory - OpenCV
- Scale down to reduce further computations
- Histogram equalization in HSV color space
- Low-pass filtration using convolution in RGB space
- Final conversion to HSV
- Separate thresholding for every color in the logo (red, green, blue, yellow)
- Segment detection using the flood fill algorithm. Calculation of descriptive features per segment.
- (Optional) Closing (dilation followed by erosion). Recalculation of descriptive features if necessary.
- Segments filtration based on descriptive features.
- Identification using topological correspondence.
The project was aimed at practicing and implementing typical operations in the image recognition processing pipeline. Accordingly, the implementation is rather straight-forward.
