2. Describe a simple algorithm to remove salt and pepper noise from a monochrome image.`
The work for this can be found in filters.py. The included routines perform a simple convolution of a median or weighted median kernel over the input.
3. Describe a simple method to perform auto white balance (AWB) on an RGB image.
The work for this can be found in awb.py. The included routines perform white balance assuming the Gray world approximation. The von Kries transform method is used to scale the LMS response to a specified illuminant (a CIE D65 illuminant is used as an example).
4. Describe a simple method to convert a Bayer image directly into monochrome.
The work for this can be found in both demosaicing.py and convert.py. The included routines perform simple NN interpolation and weighted averaging.
- numpy
- cv2
Each class (filters.py, awb.py and demosaicing.py) has its own main function. Arguments can be found by specifying the --help flag on the command line.
There is also a script run_test.py, which can execute each processing stage sequentially, e.g.
$ python run_test.py --n --d --a --f --v
Input (credit: https://github.com/codeplaysoftware/visioncpp/):
Adding artificial s&p noise:
Demosaiced to RGB:
Auto-white balanced:
Converted from RGB to monochrome:
Median filtered:
