How to recolor image object using mask?

Question

Trying to recolor the hair using a mask. Firstly segmented hair from the main image & trying to make it a realistic one changing HSV value but according to my code the result is not the accurate output that i am looking for. Any solution?

img = cv2.imread('model-demo.jpg')
img = cv2.resize(img, (256, 256))
_, mask_hair = cv2.threshold(mask_hair, thresh=210, maxval=255, type=cv2.THRESH_BINARY)

#cutting specific portion of hair from image
cut_mask = np.invert(mask_hair)
res = np.where(cut_mask, 0, img)

#HSV value changing for new color
hsv = cv2.cvtColor(res, cv2.COLOR_BGR2HSV)
hsv[:,:,0] +=120
hsv[:,:,1] +=60
hsv[:,:,2] -=20
hsv = cv2.cvtColor(res, cv2.COLOR_HSV2RGB)

plt.imshow(hsv)

The image I have:

The result according to my code:

The result I want:

PART- 2

Also tried to multiply the image mask with color but end the end it doesn't even matter...

img = cv2.resize(cv2.imread('model-demo.jpg'),(256,256))
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
mask = cv2.resize(cv2.imread('./mask.jpg'),(256,256))

mask_clr = np.zeros([256,256,3],dtype=np.uint8)
mask_clr[np.where((mask_hair==255).all(axis = 2))] = [0, 255, 0]

imgMultiply = cv2.multiply(img,mask_clr)

mask_clr --> green mask:

You add/subtract colours with+= or -=. I believe that you're getting an overflow of integer values (e.g. adding 120 causes to go over 255). — decadenza
– decadenza, Commented Jun 12, 2021 at 10:31
@decadenza Thanks for your reply. As per your suggestion, I tried changing the value within the range but nothing happened ! can you check my PART-2 (edited code above). Where I tried with mask overlay for changing color like realistic but nothing happened !! — MSI
– MSI, Commented Jun 12, 2021 at 12:08

Abdelsalam Hamdi · Accepted Answer · 2021-06-13 03:03:46Z

2

I have done a simple program for your project

first I used this code to get a mask https://docs.opencv.org/3.4/da/d97/tutorial_threshold_inRange.html . This code uses inRange function.

after that I used the following code to get the results

a = np.where(frame_threshold > 0)
ones = np.ones_like(img)
ones[a] = [0.5,1,0.5]
r = img*ones

The results I got

if you change the values to be [0.3,1,1] it will be yellow.

and it will be red if the value is [0.3,0.1,1]

and it will be green bluish if the value is [1.5,1.0,0.8]

edited Jun 13, 2021 at 3:03

answered Jun 12, 2021 at 13:42

Abdelsalam Hamdi

5254 silver badges18 bronze badges

Sign up to request clarification or add additional context in comments.

9 Comments

MSI Over a year ago

@abbs I appreciate your answer as well as suggestions. Can you please elaborate me these values -> [0.5,1,0.5] ? For mixing any color with rgb is it the right approach?

MSI Over a year ago

Another thing with this line i got error in my code I hade to change it into ones[np.where((mask_hair>0).all(axis = 2))] = [0.5,1,0.5]

Abdelsalam Hamdi Over a year ago

@MSI basically I have factor for (RGB) values. so if I want the color to be green the green value must be the greatest. So if I multiply all R and B by 0.5 and G by one the color will become more green. Hope that answered your question. I would appreciate it if you can vote for my answer if you find my answer helpful.

MSI Over a year ago

Yes, it was! just a little inquiry. If I want it to turn in yellow is it possible?

Abdelsalam Hamdi Over a year ago

basically the factors are for [B,G,R] so the dominant color shall be the greatest. see the examples I added

|

Collectives™ on Stack Overflow

How to recolor image object using mask?

1 Answer 1

9 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

9 Comments

Your Answer

Sign up or log in

Post as a guest

Related