Retinex implementation python

Asked Aug 17 '20 at 16:57

Active Nov 10 '20 at 17:13

Viewed 931 times

I am not really new to python programming, but I am just confused about implementing fundamental retinex (Single Scale Retinex).
I know the Retinex formula : retinex = logI(x,y) - (log(F(x,y)) * I(x,y)) and here is my code:

def singleScaleRetinex(img, sigma):
    retinex = np.log10(img) - np.log10(cv2.GaussianBlur(img, (0,0), sigma))
return retinex

After running my code for single scale output image was almost going to dark just like edge detection. I did on a higher scale and lower scale, but the same output.

Can you help my problem ?
is the code wrong ?

add some question:
we now that

logR(x,y) = LogI(x,y) - Log[G(x,y)*I(x,y)]

but why R(x,y) = LogR(x,y) ? (explain me)

here is my output

edited Nov 10 '20 at 17:13

SecretAgentMan

2,856
7
21
41

asked Aug 17 '20 at 16:57

anthonius adi

What did you expect as output? If you take the log of the image, you’ll have lower values. If you then subtract the local mean, you’ll have even lower values. – Cris Luengo Aug 17 '20 at 17:45
I expected that output of single scale retinex were better than original image. – anthonius adi Aug 18 '20 at 06:59
It solved anyway. – anthonius adi Oct 08 '20 at 17:25

Retinex implementation python

0 Answers0