Warning: Can not open [/home/conf/public_html/cgi-bin/show_python.log]. Ignore cimg = img.copy() if '_x' in ftype: # filter[:, iy0] = cx filter = np.zeros([ksize[0], 1]) filter[:, 0] = cx cimg = cv2.filter2D(cimg, -1, filter) if '_y' in ftype: # filter[ix0, :] = cy filter = np.zeros([1, ksize[1]]) filter[0, :] = cy cimg = cv2.filter2D(cimg, -1, cy) Download script from image_contrast.py Related files: image_contrast.gif import sys import numpy as np from tqdm import tqdm, trange from scipy.signal import savgol_coeffs import cv2 import matplotlib.pyplot as plt infile = 'stem.png' outfile = 'normalized_image.png' block_w = 20 block_h = 10 copy_w = 3 copy_h = 3 smooth_mode = 'poly' # 'gauss' # for gauss and poly smooth_w = 15 smooth_h = 15 # for poly smooth_order = 3 view_w = 800 def usage(): print() print(f"Usage: python {sys.argv[0]} infile block_w block_h copy_w copy_h view_w") def initialize(): global infile, outfile, block_w, block_h, copy_w, copy_h global smooth_mode, smooth_w, smooth_h, smooth_order global view_w infile = 'stem.png' outfile = 'normalized_image.png' block_w = 20 block_h = 10 copy_w = 3 copy_h = 3 smooth_mode = 'poly' smooth_w = 15 smooth_h = 15 smooth_order = 3 view_w = 800 def update_vars(): global infile, outfile, block_w, block_h, copy_w, copy_h global smooth_mode, smooth_w, smooth_h, smooth_order global view_w nargv = len(sys.argv) if nargv >= 2: infile = sys.argv[1] if nargv >= 3: block_w = int(sys.argv[2]) if nargv >= 4: block_h = int(sys.argv[3]) if nargv >= 5: copy_w = int(sys.argv[4]) if nargv >= 6: copy_h = int(sys.argv[5]) if nargv >= 7: smooth_mode = sys.argv[6] if nargv >= 8: smooth_w = int(sys.argv[7]) if nargv >= 9: smooth_h = int(sys.argv[8]) if nargv >= 10: smooth_order = int(sys.argv[9]) if nargv >= 11: view_w = int(sys.argv[10]) def read_image(infile, color = cv2.IMREAD_GRAYSCALE): print() print(f"Read image [{infile}]") return cv2.imread(infile, color) def save_image(img, outfile): print() print(f"Save converted image to [{outfile}]") cv2.imwrite(outfile, img) def plot_image(img1, img2, img3, view_w = 600, mode = 'cv2'): print() print("plot") if len(img1.shape) == 2: height, width = img1.shape else: height, width, _ = img1.shape print(f"image shape: {width} x {height}") mag = view_w / width print(f"view width: {view_w} = {width} * {mag:.3f}") if img1 is not None: img1_resized = cv2.resize(img1, (int(width * mag + 1.0e-5), int(height * mag + 1.0e-5)), interpolation=cv2.INTER_LINEAR) if img2 is not None: img2_resized = cv2.resize(img2, (int(width * mag + 1.0e-5), int(height * mag + 1.0e-5)), interpolation=cv2.INTER_LINEAR) if img3 is not None: img3_resized = cv2.resize(img3, (int(width * mag + 1.0e-5), int(height * mag + 1.0e-5)), interpolation=cv2.INTER_LINEAR) if mode == 'cv2': cv2.imshow('Original Image', img1_resized) cv2.imshow('Converted Image', img2_resized) cv2.waitKey(0) cv2.destroyAllWindows() else: plt.figure(figsize=(15, 5)) if img1 is not None: plt.subplot(1, 3, 1) plt.title('Original Image') plt.imshow(cv2.cvtColor(img1_resized, cv2.COLOR_BGR2RGB)) if img2 is not None: plt.subplot(1, 3, 2) plt.title('Smoothed Image') plt.imshow(cv2.cvtColor(img2_resized, cv2.COLOR_BGR2RGB)) if img3 is not None: plt.subplot(1, 3, 3) plt.title('Converted Image') plt.imshow(cv2.cvtColor(img3_resized, cv2.COLOR_BGR2RGB)) plt.pause(1.0e-5) input("\nPress ENTER to terminate>>\n") def smoothing(img, ftype, ksize, sigmaX = -1, norder = 3): if 'gauss' in ftype: print("gaussian smoothing:") return cv2.GaussianBlur(img, ksize, sigmaX) elif ftype == 'fft': print("fft smoothing:") fted = np.fft.fft2(img) fted_shift = np.fft.fftshift(fted) rows, cols = img.shape cy, cx = rows // 2 , cols // 2 ry, rx = ksize for iy in range(rows): for ix in range(cols): # r2 = ((ix - cx) / rx)**2 + ((iy - cy) / ry)**2 # if r2 > 1.0: if iy < cy - ry or cy + ry < iy or ix < cx - rx or cx + rx < ix: fted_shift[iy][ix] = 0.0 fted_origin = np.fft.ifftshift(fted_shift) img_back = np.fft.ifft2(fted_origin) img_back = np.abs(img_back) # img_back = img_back / max(img_back.flatten()) * 255.0 img_back8 = np.zeros([rows, cols], np.uint8) for iy in range(rows): for ix in range(cols): img_back8[iy][ix] = img_back[iy][ix].real # return img_back.astype(np.uint8) return img_back8 else: print(f"{norder}-th order polynomial fit smoothing using {ksize[0]} x {ksize[1]} points:") # filter = np.array([[1, 1, 1], [1, 1, 1], [1, 1, 1]]) / 9.0 cx = savgol_coeffs(ksize[0], norder) cy = savgol_coeffs(ksize[1], norder) filter = np.zeros(ksize) ix0 = int(ksize[0] / 2) + 1 iy0 = int(ksize[1] / 2) + 1 filter[:, iy0] = cx / cx[ix0] filter[ix0, :] = cy / cy[iy0] tot = sum(sum(row) for row in filter) filter /= tot cimg = cv2.filter2D(img, -1, filter) ''' cimg = img.copy() if '_x' in ftype: # filter[:, iy0] = cx filter = np.zeros([ksize[0], 1]) filter[:, 0] = cx cimg = cv2.filter2D(cimg, -1, filter) if '_y' in ftype: # filter[ix0, :] = cy filter = np.zeros([1, ksize[1]]) filter[0, :] = cy cimg = cv2.filter2D(cimg, -1, cy) ''' return cimg def normalize_image(img, block_w, block_h, copy_w = 3, copy_h = 3): hdx0 = int(block_w / 2) hdy0 = int(block_h / 2) hcx0 = int(copy_w / 2) hcy0 = int(copy_h / 2) img_normalized = img.copy() height, width = img.shape def get_range(i, imax, w0): if i == 0 or i == imax - 1: w = 1 elif i < w0: w = i elif imax - 1 - w0 < i: w = imax - w0 else: w = w0 range0 = i - w range1 = i + w if range0 < 0: range0 = 0 if imax <= range1: range1 = imax return w, range0, range1 for yc in trange(0, height + 2 * hcy0, 2 * hcy0 + 1): hdy, norm_y0, norm_y1 = get_range(yc, height, hdy0) for xc in range(0, width + 2 * hcx0, 2 * hcx0 + 1): hdx, norm_x0, norm_x1 = get_range(xc, width, hdx0) # print(f"xc/{width}=", xc, hdx, norm_x0, norm_x1) copy_y0 = yc - hcy0 copy_y1 = yc + hcy0 copy_x0 = xc - hcx0 copy_x1 = xc + hcx0 # print("copy:", copy_x0, copy_x1) block = img[norm_y0:norm_y1 + 1, norm_x0:norm_x1 + 1] vmin = block.min() vmax = block.max() # print("vmin,vmax=", vmin, vmax) if vmin != vmax: img_normalized[copy_y0:copy_y1 + 1, copy_x0:copy_x1 + 1] \ = (img[copy_y0:copy_y1 + 1, copy_x0:copy_x1 + 1] - vmin) / (vmax - vmin) * 255.0 return img_normalized def main(): global infile, outfile, block_w, block_h, copy_w, copy_h, view_w, smooth_w, smooth_h initialize() update_vars() img = read_image(infile) height, width = img.shape print("image shape:", height, width) print("color range:", img.min(), img.max()) print("block size:") print(" normalize:", block_w, block_h) if copy_w > block_w: copy_w = block_w if copy_h > block_h: copy_h = block_h print(" copy :", copy_w, copy_h) smooth_w = 2 * int(smooth_w / 2) + 1 smooth_h = 2 * int(smooth_h / 2) + 1 print(f"smoothing block ({smooth_mode}): {smooth_w} x {smooth_h}") img_smoothed = smoothing(img, smooth_mode, (smooth_w, smooth_h), -1, smooth_order) img_normalized = normalize_image(img_smoothed, block_w, block_h) save_image(img_normalized, outfile) plot_image(img, img_smoothed, img_normalized, view_w, mode = 'matplotlib') # plot_image(img, ftr, fti, mode = 'matplotlib') usage() if __name__ == '__main__': main()