import gradio
import numpy as np


#import matplotlib.pyplot as plt
#from skimage import morphology,measure,feature
#from skimage.measure import label

import skimage

from skimage import restoration
from skimage.filters import threshold_otsu, rank
from skimage.morphology import closing, square, disk


def inference(img):
  gray = skimage.color.rgb2gray(img) # gray
  #binarized = np.where(grayscale>0.1, 1, 0)
  #processed = morphology.remove_small_objects(grayscale.astype(bool), min_size=33, connectivity=4).astype(int)
    
  #out = morphology.remove_small_objects(out , min_size=2, connectivity=4)
  #out = morphology.remove_small_holes(out , min_size=2, connectivity=4)

  #out = processed
  #edges = get_edges(img.copy())
  #edges = feature.canny(out, sigma=3) # edge detect via canny with sigma 3 
  #out = morphology.remove_small_objects(label(edges), 2,) # noise_reduced
  #out = morphology.remove_small_objects( out , 2,) # noise_reduced


  # black out pixels
  #mask_x, mask_y = np.where(processed == 0)
  #img[mask_x, mask_y, :3] = 0
  #mask_x, mask_y = np.where(processed == 0)
  #im[mask_x, mask_y, :3] = 0

  #denoise = restoration.denoise_tv_chambolle( out , weight=0.1)
  #thresh = threshold_otsu(denoise)
  #thresh = threshold_otsu(gray)
  thresh = 0.4
  #out = closing(denoise  > thresh, square(2))
  out =gray>thresh


  return out

# For information on Interfaces, head to https://gradio.app/docs/
# For user guides, head to https://gradio.app/guides/
# For Spaces usage, head to https://huggingface.co/docs/hub/spaces
iface = gradio.Interface(
  fn=inference,
  inputs='image',
  outputs='image',
  title='Noise Removal w skimage', 
  description='Remove Noise with skimage.morphology!',
  examples=["detail_with_lines_and_noise.jpg", "lama.webp", "dT4KW.png"])  
  #examples=["detail_with_lines_and_noise.jpg", "lama.webp", "test_lines.jpg","llama.jpg", "dT4KW.png"])  

iface.launch()