How to apply bilateral filter

[gptshubham595]

I have got ans(just image) for this bilateral filter but confused with Call Library Function Node

How to attempt a bilateral filter over a 2d Matrix image

[gptshubham595]

I want to apply bilateral filter to this image 

I have code for matlab how to convert it in labview

Img=360x360

patchVar = std2(img)^2;

DoS = 2*patchVar;

J = imbilatfilt(I,DoS);

imshow(J)

[GerdW]

No need for double posts, keep things in one place...

[gptshubham595]

I have a python code also for this without any  builtin function 

But unable to convert this for Labview

import numpy as np
import cv2
import sys
import math


def distance(x, y, i, j):
    return np.sqrt((x-i)**2 + (y-j)**2)


def gaussian(x, sigma):
    return (1.0 / (2 * math.pi * (sigma ** 2))) * math.exp(- (x ** 2) / (2 * sigma ** 2))


def apply_bilateral_filter(source, filtered_image, x, y, diameter, sigma_i, sigma_s):
    hl = diameter/2
    i_filtered = 0
    Wp = 0
    i = 0
    while i < diameter:
        j = 0
        while j < diameter:
            neighbour_x = x - (hl - i)
            neighbour_y = y - (hl - j)
            if neighbour_x >= len(source):
                neighbour_x -= len(source)
            if neighbour_y >= len(source[0]):
                neighbour_y -= len(source[0])
            gi = gaussian(source[neighbour_x][neighbour_y] - source[x][y], sigma_i)
            gs = gaussian(distance(neighbour_x, neighbour_y, x, y), sigma_s)
            w = gi * gs
            i_filtered += source[neighbour_x][neighbour_y] * w
            Wp += w
            j += 1
        i += 1
    i_filtered = i_filtered / Wp
    filtered_image[x][y] = int(round(i_filtered))


def bilateral_filter_own(source, filter_diameter, sigma_i, sigma_s):
    filtered_image = np.zeros(source.shape)

    i = 0
    while i < len(source):
        j = 0
        while j < len(source[0]):
            apply_bilateral_filter(source, filtered_image, i, j, filter_diameter, sigma_i, sigma_s)
            j += 1
        i += 1
    return filtered_image


if __name__ == "__main__":
    src=cv2.imread(str(sys.argv[1]), 0)
    filtered_image_own = bilateral_filter_own(src, 5, 12.0, 16.0)
    cv2.imwrite("filtered_image_own.png", filtered_image_own)

[gptshubham595 · Accepted solution]

finally solved this

Thank you @Raven, @GerdW, and all community members 

many times I was unclear with my posts coz I was breaking parts of this in a simpler form, I failed in some attempts to explain clearly, after learning from various other docs and with community help! Successfully got this!

Thank You!

will improve me! 😇

[GerdW]

Hi gpt,

---

@gptshubham595 wrote:

finally solved this

Successfully got this!

---

Well, are you sure you need those two innermost loops?

Why are there so many coercion dots, especially on your own subVIs?

Still need to cleanup the block diagram! (Like the hidden wire from outermost loop "i" terminal to tunnel of next loop…)

Use better labels for controls/indicators. Using "integer" for an U32 indicator seems senseless…

[gptshubham595]

There are a total of 4 loops in python so yes

I am using double+int so I think it is showing red coercion dots

Sure, will try to clean up and make it look better in term of displaying some hidden wires

For labels it was a test for some reason so text not mattered much as of now renamed it btw!

Thank You!

[Rahulbala]

Can you attach the code for it?

Thanks

Rahul

[gptshubham595]

...

[gptshubham595]

Hey Rahul you can use this directly to do this using above python code and python integration toolbox via Enthought