Web service allows you to work in two modes. The first mode allows you to study the quality of filtering by different methods (median filtering, filtering by the author's method of least finite differences (MLFD) or modified method - MMLFD) images, in which the user can add different types of noise (pulse noise, grayscale noise) and different intensities (noise filling from 0% to 100%). The second mode allows you to filter images of already noisy photos that can be downloaded from external sources using the "Select File" button at the top of the main page.
Developed Web Service Interface allows you to select images from a list or download an arbitrary image file to perform filtering procedures. To do this, you must:
1. Select a file to processing: - download from a repository that already has the appropriate files, or use a download from a local disk (or local storage, for example for smartphones).
2. Go to the "Image noise" field. To work in the first mode, the user can add additive noise of the following types: "Salt" - random emissions of maximum brightness - white dots (choose the percentage of noise - from 0 to 100%); "Pepper" - random emissions of minimum brightness - black dots (choose the percentage of noise - from 0 to 100%); Gray Grades - Random Grayscale Emissions (Choose Noise Fill Percentage - 0 to 100%). After setting the values, press the appropriate button.
3. Provides the ability to create noise for a color image ( Noise option), Sepia image style. After such manipulations, a noisy black and white image is formed, which is subject to filtering.
4. Depending on the noise generated or present in the image (when loading a noisy image), go to the field Image filtering with noise type "Salt" , "Image filtering with noise type " Pepper " or Grayscale noise filtering
5. Upload an image to the appropriate field (click the "Upload" button) or delete it ("Clear" button). p>
6. To implement classic median filtering, press the buttons: (Median filter "4") (implemented filter type "Median-4") or (Median filter "8") (filter type "Median-8" is implemented) to be able to compare with the results of filtering when using known software applications (for example, Adobe PhotoShop, GIPM ).
7. For the implementation of author's filtering by the method of least finite differences (MLFD) set the data input field from 1 to 255 (number of iterations over one pixel of the image) and press the buttons: (MLFD 1) (one-time algorithm of the MLFD method is implemented) or (MLFD 3) (the algorithm of the MLFD method is implemented three times). Additionally, the user can install the following filtering parameters for MLFD, as the number of neighboring pixels to be processed (filter core). The default number is "8" (recommended), but it is possible to set to "4" (For some types of noisy images, it gives better results than for "8").
8. The filtering result can be observed by the user in the corresponding window on the main field after the "Progress Bar" appears, as well as the quality of the filtering by the criterion of standard deviation "SLE" in the Filtering efficiency field
9. To implement additive Gaussian noise filtering (an example of such an image is
download ready-made option on the main menu - the topmost option of the list located on the main page)
the modified method of MLFD is applied - it is caused when pressing the MLFD button.
For the algorithm to work on this advanced method, the user must set the following parameters:
- amplitude values in the range -255 to +255;
- number of adjacent filter pixels: "4" or "8". "8" is selected by default;
- depth of filtering - the number of iterations during which they occur
change the pixel values of the image. Set the multiplier n from 1 to 15, which means n * 10 ^ 6 times);
- filter character, which can be "+", "-", "+/-" -
gradient of random pixel value change during filtering. The default is set to "-".
For example, when filtering noise such as "Granular white noise", it is recommended to reduce the brightness
white spots, so the gradient will be "-", if there are black spots in the image - choose "+",
if there are white and black spots - choose "+/-".