I wrote the code review for 2 more plug-ins: Cartoon and Photocopy

Gimp dialog function do all almost the same thing: Let the user choose the
parameters for each plugin by opening a box with a preview box. While
changing the parameters it changes the preview image until the user press
the ok or the cancel button.
We can say the same thing for run function (almost the same). It gets the
drawable in the drawable structure the sets the tile cache size, gets data
from the keyboard and run the dialog, if there is no dialog it initiates its
own values for the plug-in. Checks if everything is all right, run the
plugin and stores the data.

1. Cartoon (in gegl the base class would be AREA FILTER):
    Algorithm:
    For each pixel it calculates the pixel intensity by comparing the pixels
relative intesity to its neighborhood pixels and to the relative intensity
difference to total black .
Let say mask radius is equal with radius of pixel neighborhood for intensity
comparison, threshold is the relative intensity difference which will result
in darkening, ramp is the amount of relative intensity difference before
total black and blur radius is mask radius per 3.
Then the new intensity of the pixel will be:
 relative difference = pixel intensity / average (mask radius)
 If relative difference < Threshold
 intensity multiply = (Ramp - MIN (Ramp, (Threshold - relative difference)))
/ Ramp
 pixel intensity =old intensity * intensity mult

    static void cartoon:
    Checks for the preview, then sets the width and height of the drawable
image, gets the image type (bytes) and the aplha value (has_alpha). It
initialize the 5 vectors and 2 destination image structures (dest1 for blur
radius and dest2 for mask radius). Calculates the standard deviations from
blur and mask radius. Then derives the constant values for calculating the
gaussian's from the deviations (via the 4th order approximation of the
gaussian operator).
Like in the case of gaussian blur the calculation of the new values of the
image is linear so the calculation can be devided for 2 directions. First
calculates the values for every column then for every row.
Calculating for every column: firstly initializes and calculates the first
and the last pixel of the column. Then with the help of the gaussian
constants it calculates every pixel of the column the transfers the pixels
to the destination image.
It will do the same calculations in case of the horizontal direction.
After calculating the blakc percentage value (ramp). Then calculates the new
intensity for each pixel:
relative difference = pixel intensity / average (mask radius)
 intensity multiply=1
 If relative difference < Threshold
 intensity multiply = (Ramp - MIN (Ramp, (Threshold - relative
difference))) / Ramp
 pixel intensity =old intensity * intensity multiply
 Before upgrading the drawable image transfers the calculated destination
image structure from RGB format to HLS, sets the lightness and converts
back.

    computer ramp:
    Calculates the ramp value (intensity difference from total black) by
calculating the difference between the destination images (one calculated
with blur radius the other with mask radius), and hysterizes the difference.
Then compares the hysterized values to the percentage of the black color and
calculates the relative intensity via average.


    2. Photocopy (in gegl the base class would be AREA FILTER):
    Propagates dark value in an image based on each pixel's relative
darkness to a neighboring average. Sets the remaining pixels to white.
    The plug-in differs a little from the cartoon plug-in.
    Algorithm:
    Using the same notations as in the cartoon plug-in the new intensity of
every pixel will be:
elative diff = pixel intensity / avg (mask radius)
 If relative diff < Threshold
    intensity mult = (Ramp - MIN (Ramp, (Threshold - relative diff))) / Ramp
    pixel intensity *= intensity mult
 Else pixel intensity = white

    static void photocopy: It is almost the same as in the cartoon plug-in.
    Desaturates the image, checks for the preview, then sets the width and
height of the drawable image, gets the image type (bytes) and the aplha
value (has_alpha). It initialize the 5 vectors and 2 destination image
structures (dest1 for blur radius and dest2 for mask radius). Calculates the
standard deviations from blur and mask radius. Then derives the constant
values for calculating the gaussian's from the deviations (via the 4th order
approximation of the gaussian operator).
Like in the case of gaussian blur the calculation of the new values of the
image is linear so the calculation can be devided for 2 directions. First
calculates the values for every column then for every row.
Calculating for every column: firstly initializes and calculates the first
and the last pixel of the column. Then with the help of the gaussian
constants it calculates every pixel of the column the transfers the pixels
to the destination image.
It will do the same calculations in case of the horizontal direction.
After calculating the black and white percentage values (ramp down and ramp
up). Relative difference is equal with the average pixel intensity of the
blur radius image per the average pixel intersity of the maske radius image.
While calculating the intensity of each pixel it also calculates the
lightness of the drawable image between 0 and 255.
   intensity multiply = 1
   if relative diff < Threshold
    if ramp_down == 0 intensity multiply = 0
        else intensity mult = (ramp_down - MIN (ramp_down, (Threshold -
relative
difference))) / ramp_down
   else if ramp_down !=0 mult = MIN (ramp_up,(relative difference -
Threshold)) / ramp_up;
    Knowing the lightness calculates the value of each pixel and draws the
image.

    compute ramp and find constants functions are the same as in the cartoon
plug-in.



    Compute ramp and find constants could be directly written in gegl api as
not to write the same thing twice. For recalculating the pixels by adding
lightness we could use a lightness plug-in for it.

    I have an exam saturday, so I will finish and send the plug-in ported to
gegl on sunday.
    If there is anything else I should do please write it. If there is
something wrong with the code reviews tell it and I will rewrite it.

    Thanks,
    Robert Sasu
_______________________________________________
Gimp-developer mailing list
Gimp-developer@lists.XCF.Berkeley.EDU
https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer

Reply via email to