Author: Antony Cherepanov
E-mail: antony.cherepanov@gmail.com
FFilter is an app that can be used for studing different types of noise generators / noise filters in image processing. Currently only filtration of grayscale images are supported.
Tested on Ubuntu 12.04 with gcc 4.6.3, Qt 4.8 and higher.
cd /path/to/folder/with/FFilter
qmake -r
makeMain window of the app consists of the following elements (downward from left to right):
- menu bar
- three image labels for original, noisy and filtered images
- button "Open image" for loading original image
- buttons for saving noisy and filtered images
- slider for setting noise level
- label that show standard deviation value
- button "Filter image" that start filtration process
- progress bar
First of all, you need to load an image to which you will be working. Click "Open image" button and choose image. This image will be loaded in all three image labels. If you pick color image, FFilter will transform it to greyscale image.
Next you can set noise settings. Go to Menu -> Noise -> Noise settings. In opened window "Noise Settings" you can choose noise type and noise amplitude (AMP):
- if you choose Constant noise type, then algorithm will choose N pixels and add to their luminance AMP value
- if you choose Random noise type, then algorithm will choose N pixels and add to their luminance random value within [0, AMP]
- if you choose Absolute Random noise type, then algorithm will choose N pixels and add to their luminance some random value within [0, MAX_LUM]
With the slider "Noise level" you can choose how many pixels (in percents) will be noised. The result noisy image will be set to the central (second) image label.
Let's move to filtration section. Go to Menu -> Filter -> Filter mask settings. In opened window "Mask settings" you can set mask parameters. Mask is a collection of pixels, that are involved in filtration process. Each mask must have one central pixel (colored red in mask) - this is the pixel that is currently processing. Also mask can contain other pixels - neighbors of the central pixel. Also each mask pixel can have "weight". This value helps filtration algorithm to understand importance of the pixel.
In "Mask settings" dialog you can set the size of the mask (number of rows and columns). By clicking the right mouse button on the mask cell you will call its' context menu. In context menu you can enable/disable pixel or set it as central pixel. To set "weight" for the enabled pixel, click the left mouse button on the mask cell and edit its value.
Example mask: set mask size 3 x 3. Central pixel - pixel with coordinates [2, 2]. Enable all pixels in mask and set for all of them weight "1".
To set filtration method, go to Menu -> Filter -> Aggregation filter settings. In opened dialog "Filter settings" you can set type of aggregation operator. What is aggregation operator? This is the function that gives the filtered value for the pixel. The calculation is made on the basis of mask parameters. Let's consider the available operators:
- Min - this operator will choose mask pixel with minimum value (luminance)
- Max - this operator will choose mask pixel with maximum value
- Mean - this operator will calc mean value of the mask pixels. For this aggregation operator you can set "power" parameter. Value of each mask pixel will be raised to the defined power before final calculation of mean value.
- Geometric Mean - this operator will calc geometric mean value of the mask pixels
- Harmonic Mean - this operator will calc harmonic mean value of the mask pixels
- Medain - this operator will choose mask pixel with mean value
- Functional - this operator will calc function (natural logarithm or exponent) for each mask pixel, summ all results and then calc inverse function
After you setup all parameters, you can start filtration process by clicking on "Filter image" button. Filtered image will be set on the right (third) image label.