insituTEM 0.1.6__py3-none-any.whl → 0.1.7__py3-none-any.whl

insituTEM/insitu_Diff.py

@@ -0,0 +1,359 @@
+"""
+in-situ TEM Toolbox - Diffraction pattern analysis
+
+Assembles of functions related to process FFT and diffraction patterns
+
+Example:
+    
+    import insitu_Diff as Diff  
+    Diff.f2tif(path,1)
+
+Created on Aug 8 2024
+@author: Meng Li
+"""
+import numpy as np
+import cv2
+# from skimage.feature import blob_dog
+import matplotlib.pyplot as plt
+# from scipy import ndimage
+import math
+
+# import pandas as pd
+
+def getFFT(im):
+    """
+    Function to calculate FFT of the image
+    input: im- np array of image data
+    output: fft image
+    """
+    # Get the original dimensions of the image
+    h, w = im.shape
+
+    size = min(h,w) #in case h w don't match
+    im_crop=im[:size,:size] #crop image 
+    fft_image = np.fft.fft2(im_crop)
+    fft_shifted = np.fft.fftshift(fft_image)
+    fft = 20 * np.log(np.abs(fft_shifted) + 1e-10)  # Add small constant to avoid log(0)
+    
+    return fft
+
+def spotFFT(fft, sigmaMax=2.5, sigmaMin=2.45, sigma_ratio=20, thrs=0.98, R_tol=2, fftfilter=3, tolerance=5, display=True):
+    """
+    Function to get diffraction spot positions in FFT
+    Input: fft- FFT image
+            sigmaMax, sigmaMin, sigma_ratio, thrs - Parameters for blob detection
+            R_tol - tolerance of center spot
+            fftfilter- filtersize to denoise FFT to make diffraction spots more visible
+            tolerance - tolerance for symmetry matching
+    
+    Output: 
+        Data of all detected spots:    X(px)   Y(px) 
+        Image labeling the spots
+     """
+    from skimage.feature import blob_dog
+    import matplotlib.pyplot as plt
+    from scipy import ndimage
+    import math
+  
+    import numpy as np
+    import pandas as pd
+
+    h, w = fft.shape
+    size = max(h, w)
+    scale_fft = 1 / (size)
+    
+    x_c, y_c = int(w / 2), int(h / 2)
+    image = ndimage.maximum_filter(fft, size=fftfilter, mode='wrap')
+    blobs_dog = blob_dog(image, max_sigma=sigmaMax, min_sigma=sigmaMin, threshold=thrs, sigma_ratio=sigma_ratio, exclude_border=True)
+    blobs_dog[:, 2] *= math.sqrt(2)
+    
+
+    
+    data = []
+    blob_set = set()
+    
+    for j in range(len(blobs_dog)):
+        y, x = blobs_dog[j, 1] - y_c, blobs_dog[j, 0] - x_c
+        distance = math.sqrt(x**2 + y**2)
+        if distance < R_tol:
+            continue
+        
+        data.append([blobs_dog[j, 1], blobs_dog[j, 0]])
+        blob_set.add((blobs_dog[j, 1], blobs_dog[j, 0]))
+    
+    # Ensure symmetry
+    for x, y in data:
+        sym_x = 2 * x_c - x
+        sym_y = 2 * y_c - y
+        
+        if not any(math.isclose(sym_x, bx, abs_tol=tolerance) and math.isclose(sym_y, by, abs_tol=tolerance) for bx, by in blob_set):
+            data.append([sym_x, sym_y])
+            blob_set.add((sym_x, sym_y))
+    
+    data = np.array(data)
+
+    if display == True:
+        fig, axes = plt.subplots(1, 2, figsize=(10, 5), sharex=True, sharey=True)
+        ax = axes.ravel()
+        ax[0].set_title('FFT')
+        ax[0].imshow(fft)
+        ax[1].set_title('DP detection')
+        ax[1].imshow(image)
+        plt.tight_layout()
+        plt.scatter(data[:, 0], data[:, 1], marker='+', color='red')
+        for i, (x_i, y_i) in enumerate(zip(data[:, 0], data[:, 1])):
+            plt.annotate(str(i+1), (x_i, y_i), textcoords="offset points", xytext=(0,10), ha='center', color='white')
+        
+    return data
+
+def DP_mask(fft,circle_radius=10, center_rad=1,sigmaMax=2.5, sigmaMin=2.45, sigma_ratio=20, thrs=0.98, R_tol=2, fftfilter=3, tolerance=5):
+    """
+    Create a mask image with circles drawn at detected diffraction points.
+    
+    Parameters:
+        image_shape (tuple): Shape of the image (height, width).
+        points (np.ndarray): Array of detected points [[x1, y1], [x2, y2], ...].
+        circle_radius (int): Radius of the circles to draw.
+        center_rad: radius of center circle
+        sigma...: parameters for detect DP
+    
+    Returns:
+        np.ndarray: Mask image with drawn circles.
+    """
+
+    mask = np.zeros(fft.shape, dtype=np.uint8)
+    points= spotFFT(fft,sigmaMax,sigmaMin,sigma_ratio,thrs,R_tol,fftfilter,tolerance,display=False)
+    for x, y in points:
+        cv2.circle(mask, (int(x), int(y)), circle_radius, 255, thickness=-1)
+
+    if center_rad != 0:
+        center=int(fft.shape[0]/2)
+        cv2.circle(mask, (center, center), center_rad, 255, thickness=-1)
+    return mask
+
+def maskedIFFT(image,mask,mix_ratio=0.3):
+    """
+    generate masked image from FFT
+    Input: original image; mask; mix_ratio for IFFT filtering
+    Output: FFT_image,  IFFT image, IFFT filterd image
+    Example: fft_image, ifft_image,filtered_image=maskedIFFT(image,mask,mix_ratio)
+    """
+    h,w=image.shape
+
+    if h !=w:
+        d=min(h,w)
+        image=image[0:d,0:d]
+
+    # Perform FFT on the image
+    fft = cv2.dft(np.float32(image), flags=cv2.DFT_COMPLEX_OUTPUT)
+    fft_shifted = np.fft.fftshift(fft)
+    fft_image=np.log(1 + cv2.magnitude(fft_shifted[:, :, 0], fft_shifted[:, :, 1])) #using fft_image to apply mask results in empty image
+    
+    # Ensure the mask image has the same size as the FFT
+    mask_image = cv2.resize(mask,(fft_shifted.shape[1], fft_shifted.shape[0]))
+
+    # Apply the mask to the FFT
+    masked_fft_shifted = cv2.bitwise_and(fft_shifted, fft_shifted, mask=mask_image)
+
+    # Perform IFFT on the masked FFT
+    masked_fft = np.fft.ifftshift(masked_fft_shifted)
+    masked_image = cv2.idft(masked_fft)
+    masked_image = cv2.magnitude(masked_image[:, :, 0], masked_image[:, :, 1])
+    
+    #convert to 8bit images
+    fft_image=cv2.normalize(fft_image, None, 0, 255, cv2.NORM_MINMAX, cv2.CV_8U)
+    ifft_image=cv2.normalize(masked_image, None, 0, 128, cv2.NORM_MINMAX, cv2.CV_8U) #cv2.normalize(src, dst, alpha, beta, norm_type, dtype)
+    
+    # # Normalize the pixel values to the range of 0-255
+    ifft_image_8 = cv2.normalize(ifft_image, None, 0, 255, cv2.NORM_MINMAX, cv2.CV_8U)
+    
+
+    # Add IFFT to original image with mix_ratio
+    masked_image_normalized = cv2.normalize(ifft_image, None, 0, int(255*mix_ratio), cv2.NORM_MINMAX, cv2.CV_8U)
+    image_normalized = cv2.normalize(image, None, 0, int(255*(1-mix_ratio*0.5)), cv2.NORM_MINMAX, cv2.CV_8U)
+    
+    filtered_image=cv2.add(image_normalized,masked_image_normalized)
+
+
+    return fft_image, ifft_image, filtered_image
+
+
+def measureFFT(fft,path,scale,j=0,sigmaMax=2.5,sigmaMin=2.4,thrs=0.97,R_tol=2,fftfilter=2,savefig=True,displayfig=True):
+    """
+    Function to measure diffraction spots in FFT
+    Input: fft- FFT image
+            path: path to save data
+            sigmaMax, sigmaMin, thrs - Parameters for blob detection
+            R_tol - tolerance of center spot
+            
+    Output: 
+        Data of all detected spots: Point   X(px)   Y(px)  distance(px)  angle(˚)  1/d (1/nm)  d (nm)
+        Image labeling the spots
+
+    """
+    from skimage.feature import blob_dog
+    import matplotlib.pyplot as plt
+    from scipy import ndimage
+    import math
+  
+    import numpy as np
+    import pandas as pd
+
+    h,w = fft.shape
+    size=max(h,w)
+    scale_fft=1/(scale*size)
+    
+    x_c=int(w/2)
+    y_c=int(h/2)
+    center=x_c/2 #center of cropped_fft
+    
+    fft_crop=fft[x_c-int(center):x_c+int(center),y_c-int(center):y_c+int(center)]
+    image= ndimage.maximum_filter(fft_crop, size=fftfilter, mode='wrap')
+    blobs_dog = blob_dog(image, max_sigma=sigmaMax,min_sigma=sigmaMin, threshold=thrs,sigma_ratio=10,exclude_border=True)
+    blobs_dog[:, 2] = blobs_dog[:, 2] * math.sqrt(2)
+    
+    
+    title = 'DP detection'
+    title1= 'FFT_'+str(j)
+    fig, axes = plt.subplots(1, 2, figsize=(10, 5), sharex=True, sharey=True)
+    ax = axes.ravel()
+    
+    ax[0].set_title(title1)
+    ax[0].imshow(fft_crop)
+    ax[1].set_title(title)
+    ax[1].imshow(image)
+    
+    plt.tight_layout()
+    
+    pathout=path[:-4]+'_'+str(j)+'_FFT.csv'
+    pathout_img=path[:-4]+'_'+str(j)+'_DP.tif'
+    
+    
+    data=np.zeros((len(blobs_dog)-1,8))
+    
+    
+    idx=0
+    for j in range(len(blobs_dog)):
+    
+    
+        y=blobs_dog[j,1]-center
+        x=blobs_dog[j,0]-center
+    
+        
+        distance=math.sqrt(x**2+y**2)#px distance to center
+        if distance<R_tol: #skip center point
+            continue
+
+        theta=math.atan2(x,y)#atan gives +-90˚ values, atan2 gives 180
+        angle=0-math.degrees(theta)
+        data[idx,:7] = [j , blobs_dog[j, 1], blobs_dog[j, 0], distance, angle, distance * scale_fft, 1 / (distance * scale_fft)]
+        idx += 1
+    
+    
+    plt.scatter(data[:,1],data[:,2],marker='+',color='red')
+    for i, (x_i, y_i) in enumerate(zip(data[:,1],data[:,2])):
+        plt.annotate(str(i+1), (x_i, y_i), textcoords="offset points", xytext=(0,10), ha='center',color='white')
+    
+    
+    data[:,7]=data[:,4]-data[0,4]
+    
+    columns= ["Point","X(px)","Y(px)","distance(px)","angle to X(˚)","1/d (1/nm)","d (nm)","angle to P1(˚)"]
+    # Set pandas global display format for floats
+    pd.options.display.float_format = '{:.3f}'.format
+    # Create a DataFrame
+    df = pd.DataFrame(data, columns=columns)
+    df.to_csv(pathout,index=False)
+    if savefig:
+        plt.savefig(pathout_img,bbox_inches='tight', pad_inches=0.1,dpi=150)
+    if displayfig:
+        plt.show()
+
+    return df
+
+
+
+
+def FFT2Polar(fft):
+    """
+    Function to convert fft image to polar image
+    input: fft pimage 
+    output: polar FFt image
+    """
+    import cv2
+    #function to convert fft image to polar coordinates from the center from 0-180 deg
+    center = (fft.shape[1] // 2, fft.shape[0] // 2)
+    polar_fft = cv2.linearPolar(fft, center, center[0], cv2.WARP_FILL_OUTLIERS)
+    # polar_fft= polar_fft[:,:center[1]]
+    #resize the image to realistic size
+    resized_polar_fft = cv2.resize(polar_fft, (center[0], 360))
+        # plt.imshow(resized_polar_fft)
+    
+    p_fft1=resized_polar_fft[:180,:]
+    p_fft2=resized_polar_fft[180:,:]
+    p_fft=p_fft1+p_fft2
+    return p_fft
+
+def maskedFFT(image,mask):
+    """
+    generate masked image from FFT
+    Input: original image; mask
+    Output: FFT_image, filtered IFFT image
+    """
+    h,w=image.shape
+
+    if h !=w:
+        d=min(h,w)
+        image=image[0:d,0:d]
+
+    # Perform FFT on the image
+    fft = cv2.dft(np.float32(image), flags=cv2.DFT_COMPLEX_OUTPUT)
+    fft_shifted = np.fft.fftshift(fft)
+    fft_image=np.log(1 + cv2.magnitude(fft_shifted[:, :, 0], fft_shifted[:, :, 1]))
+    
+
+    # Ensure the mask image has the same size as the FFT
+    mask_image = cv2.resize(mask,(fft_shifted.shape[1], fft_shifted.shape[0]))
+
+    # Apply the mask to the FFT
+    masked_fft_shifted = cv2.bitwise_and(fft_shifted, fft_shifted, mask=mask_image)
+
+
+    # Perform inverse FFT on the masked FFT
+    masked_fft = np.fft.ifftshift(masked_fft_shifted)
+    masked_image = cv2.idft(masked_fft)
+    masked_image = cv2.magnitude(masked_image[:, :, 0], masked_image[:, :, 1])
+    
+    #convert to 8bit images
+    fft_image=cv2.normalize(fft_image, None, 0, 255, cv2.NORM_MINMAX, cv2.CV_8U)
+    masked_image=cv2.normalize(masked_image, None, 0, 128, cv2.NORM_MINMAX, cv2.CV_8U) #cv2.normalize(src, dst, alpha, beta, norm_type, dtype)
+    
+
+
+    return fft_image, masked_image
+
+
+def FFTstack(stackpath,ndfilersize=3,fps=15):
+    import tqdm
+    from scipy import ndimage
+    import tifffile
+    from insituTEM import insitu_IO as IO
+    """
+    Apply FFT to the entire stack. 
+    """
+    pathout=stackpath[:-4]+'_FFT.tif'
+    stack =tifffile.imread(stackpath)
+    
+    nFrames, h,w = stack.shape
+    print("------------------------------------------")
+    print("Getting FFT to the stack:")   
+    with tifffile.TiffWriter(pathout,bigtiff=True) as tif:
+        for i in tqdm.tqdm(range(nFrames)):
+            fr=stack[i]
+            fft=Diff.getFFT(fr)
+            if ndfilersize!=0:
+                fft = ndimage.maximum_filter(fft, size=ndfilersize, mode='wrap')
+            fft_image=cv2.normalize(fft, None, 0, 255, cv2.NORM_MINMAX, cv2.CV_8U)
+            tif.write(fft_image, contiguous=True)
+    IO.tiff2avi(pathout,fps=15)
+
+    print("Conversion done!")