Merge branch 'master' into 'master'

Master

See merge request !44

src/methods/Samanos_Thomas/Readme.md

+### How to use ?
+There are 2 functions that can be called in reconstruct.py :
+f.cfa_reconstruction_1 that uses the method 1 
+and f.cfa_reconstruction that uses the method 2
+
+They use the same arguments which are y , cfa and input_shape.

src/methods/Samanos_Thomas/function.py

+
+import numpy as np
+import matplotlib.pyplot as plt
+from src.checks import check_cfa, check_rgb 
+from src.forward_model import CFA
+import src.methods.Samanos_Thomas.version1 as v1
+import src.methods.Samanos_Thomas.version2 as v2
+from scipy import ndimage
+
+
+def cfa_reconstruction(y: np.ndarray, cfa:str, input_shape: tuple) -> np.ndarray:
+    """Performs demosaicking on y.
+
+    Args:
+        cfa (str): Name of the CFA. Can be bayer or quad_bayer.
+
+    Returns:
+        np.ndarray: Demosaicked image.
+    """
+    if cfa == "bayer":
+        op = v2.bayer_reconstruction(y,input_shape)
+    elif cfa == "quad_bayer":
+        op = v2.quad_bayer_reconstruction(y,input_shape)
+    else:
+        op = np.zeros(input_shape)
+    return op
+
+def cfa_reconstruction_1(y: np.ndarray, cfa:str, input_shape: tuple) -> np.ndarray:
+    """Performs demosaicking on y.
+
+    Args:
+        cfa (str): Name of the CFA. Can be bayer or quad_bayer.
+
+    Returns:
+        np.ndarray: Demosaicked image.
+    """
+    if cfa == "bayer":
+        op = v1.bayer_reconstruction(y,input_shape)
+    elif cfa == "quad_bayer":
+        op = v1.quad_bayer_reconstruction(y,input_shape)
+    else:
+        op = np.zeros(input_shape)
+    return op
\ No newline at end of file

src/methods/Samanos_Thomas/reconstruct.py

+"""The main file for the reconstruction.
+This file should NOT be modified except the body of the 'run_reconstruction' function.
+Students can call their functions (declared in others files of src/methods/your_name).
+"""
+
+
+import numpy as np
+
+import src.methods.Samanos_Thomas.function as f
+from src.forward_model import CFA
+
+def run_reconstruction(y: np.ndarray, cfa: str) -> np.ndarray:
+    """Performs demosaicking on y.
+
+    Args:
+        y (np.ndarray): Mosaicked image to be reconstructed.
+        cfa (str): Name of the CFA. Can be bayer or quad_bayer.
+
+    Returns:
+        np.ndarray: Demosaicked image.
+    """
+
+    input_shape = (y.shape[0], y.shape[1], 3)
+    Y = f.cfa_reconstruction(y, cfa,input_shape)
+    return Y
+
+####
+####
+####
+
+####      ####                ####        #############
+####      ######              ####      ##################
+####      ########            ####      ####################
+####      ##########          ####      ####        ########
+####      ############        ####      ####            ####
+####      ####  ########      ####      ####            ####
+####      ####    ########    ####      ####            ####
+####      ####      ########  ####      ####            ####
+####      ####  ##    ######  ####      ####          ######
+####      ####  ####      ##  ####      ####    ############
+####      ####  ######        ####      ####    ##########
+####      ####  ##########    ####      ####    ########
+####      ####      ########  ####      ####
+####      ####        ############      ####
+####      ####          ##########      ####
+####      ####            ########      ####
+####      ####              ######      ####
+
+# 2023
+# Authors: Mauro Dalla Mura and Matthieu Muller

src/methods/Samanos_Thomas/version1.py

+import numpy as np
+from skimage.transform import resize
+
+def upscale(img: np.ndarray, shape: tuple, pas: int) -> np.ndarray:
+    """Upscales the acquisition to the shape of the desired reconstruction.
+
+    Args:
+        img (np.ndarray): mosaique image in smaller size.
+        shape (tuple): Shape of the output.
+
+    Returns:
+        np.ndarray: Upscaled image.
+    """
+    return resize(img, shape, anti_aliasing=True)
+
+def bayer_reconstruction(y : np.ndarray, input_shape : tuple) -> np.ndarray:
+    op_shape = (input_shape[0]//2,input_shape[1]//2,3)
+    op = np.zeros(op_shape)
+    conv = np.zeros((2,2,3))
+    liste = [(0,1,0),(0,0,1),(1,1,1),(1,0,2)]
+    for el in liste :
+        conv[el]=1
+        if el[2]==1:
+            conv[el]=1/2
+    for i in range (3):
+        op[:,:,i] = conv2d(y,conv[:,:,i], 2)
+    op1 = upscale(op,(input_shape[0],input_shape[1],3),2)
+    return op1
+
+def quad_bayer_reconstruction(y : np.ndarray, input_shape : tuple) -> np.ndarray:
+    op_shape = (input_shape[0]//4,input_shape[1]//4,3)
+    op = np.zeros(op_shape)
+    conv = np.zeros((4,4,3))
+    liste = [(2,0,2),(3,0,2),(2,1,2),(3,1,2),(0,0,1),(1,0,1),(0,1,1),(1,1,1),(2,2,1),(2,3,1),(3,2,1),(3,3,1),(0,2,0),(0,3,0),(1,2,0),(1,3,0)]
+    for el in liste :
+        conv[el]=1/4
+        if el[2]==1:
+            conv[el]=1/8
+    for i in range (3):
+        op[:,:,i] = conv2d(y,conv[:,:,i], 4)
+    op1 = upscale(op,(input_shape[0],input_shape[1],3),4)
+    return op1
+
+def conv2d(img : np.ndarray, conv : np.ndarray, pas :int):
+    y_shape = (img.shape[0], img.shape[1])
+    op = np.zeros((y_shape[0]//pas,y_shape[1]//pas))
+    for i in range(0,y_shape[0],pas):
+        for j in range(0,y_shape[1],pas):
+            op[i//pas,j//pas] = np.sum(img[i:i+pas,j:j+pas]*conv)
+    return op

src/methods/Samanos_Thomas/version2.py

+import numpy as np
+from scipy.signal import convolve2d
+
+CONV2 = np.array([[1/4,1/2,1/4],[1/2,1,1/2],[1/4,1/2,1/4]])
+
+def bayer_reconstruction(y : np.ndarray, input_shape : tuple) -> np.ndarray:
+    op_shape = (input_shape[0],input_shape[1],3)
+    op = np.zeros(op_shape)
+    conv = np.zeros((2,2,3))
+    liste = [(0,1,0),(0,0,1),(1,1,1),(1,0,2)]
+    for el in liste :
+        conv[el]=1
+    op = masq(y,conv, 2)
+    op1 = np.zeros_like(op)
+    for color in range(3):
+        op1[:,:,color] = convolve2d(op[:,:,color],CONV2,mode ="same")
+        if color == 1 :
+            op1[:,:,color] = op1[:,:,color]/2
+    return op1
+
+CONV4 = np.array([[11/156,43/312,2/13,43/312,11/156],
+                 [43/312,1/6,1/4,1/6,43/312],
+                 [2/13,1/4,1/3,1/4,2/13],
+                 [43/312,1/6,1/4,1/6,43/312],
+                 [11/156,43/312,2/13,43/312,11/156]])
+
+def quad_bayer_reconstruction(y : np.ndarray, input_shape : tuple) -> np.ndarray:
+    op_shape = (input_shape[0],input_shape[1],3)
+    op = np.zeros(op_shape)
+    conv = np.zeros((4,4,3))
+    liste = [(2,0,2),(3,0,2),(2,1,2),(3,1,2),(0,0,1),(1,0,1),(0,1,1),(1,1,1),(2,2,1),(2,3,1),(3,2,1),(3,3,1),(0,2,0),(0,3,0),(1,2,0),(1,3,0)]
+    for el in liste :
+        conv[el]=1
+    op = masq(y,conv, 4)
+    op1 = np.zeros_like(op)
+    for color in range(3):
+        op1[:,:,color] = convolve2d(op[:,:,color],CONV4,mode ="same")
+        if color == 1 :
+            op1[:,:,color] = op1[:,:,color]/2
+    return op1
+
+def masq(img : np.ndarray,conv:np.ndarray, pas:int) ->np.ndarray:
+    masq = np.zeros((img.shape[0],img.shape[1],3))
+    po = np.zeros((img.shape[0],img.shape[1],3))
+    for color in range (3):
+        for i in range(0,po.shape[0],pas):
+            for j in range(0,po.shape[1],pas):
+                masq[i:i+pas,j:j+pas,color] = conv[:,:,color]
+        po[:,:,color] = masq[:,:,color] * img
+    return po
\ No newline at end of file