opencv python image dithering

Question

i wrote a code for image dithering in python , it is working well for some images but for some it is generating unnecessary colors in the white regions .

#IMAGE DITHERING BASED ON Floyd-Steinberg METHOD
#author : bharath kotari
#date :18-1-2018


import cv2
import numpy as np

def set_pixel(im,x,y,new):
    im[x,y]=new

def quantize(im):
    for y in range(0,height-1):
        for x in range(1,width-1):
            old_pixel=im[x,y]
            if old_pixel<127:
                new_pixel=0
            else:
                new_pixel=255
            set_pixel(im,x,y,new_pixel)
            quant_err=old_pixel-new_pixel
            set_pixel(im,x+1,y,im[x+1,y]+quant_err*w1)
            set_pixel(im,x-1,y+1, im[x-1,y+1] +  quant_err*w2 )
            set_pixel(im,x,y+1, im[x,y+1] +  quant_err * w3 )
            set_pixel(im,x+1,y+1, im[x+1,y+1] +  quant_err * w4 )


    return im

img=cv2.imread("/home/user/Downloads/blender_images/truck.jpg")
gray=cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img2=img.copy()
width,height,z=img.shape
w1=7/16.0
#print w1
w2=3/16.0
w3=5/16.0
w4=1/16.0
blue=img[:,:,0]
blue=quantize(blue)
green=img[:,:,1]
green=quantize(green)
red=img[:,:,2]
red=quantize(red)
gray1= quantize(gray)   

image = cv2.merge((blue, green, red))
cv2.imshow('original',img2)
cv2.imshow('merged',image)
cv2.imshow('gray',gray1)
cv2.waitKey(0)

attaching image for reference .. how to eliminate those big dots in the top right region ..

thankYou.

Answer 1

The spots happen because of integer overflow. cv2.imread() represents image using unsigned 8 bit integer (see numpy.uint8), which have a max value of 255 and start over from zero if that value is passed. For example: 255 + 50 = 49. This happen in your code when adding in the errors.

There are several methods to avoid this. The simplest would be to make sure your values don't go over 255.