In order to facilitate the display of user avatars on the front of the project, the avatar needs to be processed as a circle and the non-circular area as transparent. In fact, the front-end can be displayed at the time of processing, but the front-end with WebGL, temporarily uncertain, so the image from the back-end one-time processing.
So, we try to use the Linux tool Convert to complete, however, there is no solution, the subsequent decision to use Python + OpenCV.
Excellent code does not need to explain, look at the code directly, O (∩ _ ∩) O.#coding: utf8 import numpy as np import cv2 from matplotlib import pyplot as plt import glob as gb # image processing, access to the largest inscribed circle of the picture, the other area set to transparent def img_deal (input_img): # cv2.IMREAD_COLOR, read BGR Channel value, the color channel, this parameter is the function default value # cv2.IMREAD_UNCHANGED, read the alpha channel value # cv2.IMREAD_ANYDEPTH, read the gray map, the return matrix is two-dimensional img = cv2.imread (input_img, cv2.IMREAD_UNCHANGED) rows, cols, channel = img.shape # Create a 4-channel new image, including transparent channels, the initialization is transparent img_new = np.zeros ((rows, cols, 4), np.uint8) img_new [:,:, 0: 3] = img [:,:, 0: 3] # Create a single-channel image, set the maximum inscribed circle to opaque, note the coordinates of the center of the circle, cols is the x coordinate, rows is The y coordinates are: img_circle = np.zeros ((rows, cols, 1), np.uint8) img_circle [:,:,:]] = 0 # Set to full transparency img_circle = cv2.circle (img_circle, (cols / 2, rows / 2), min (rows, cols) / 2, (255), - 1) # set the maximum inscribed circle opaque # image fusion img_new #: Cv2.imencode ('. Jpg', img)  .tofile (',': ', 0] # img_circle [:,:, 0] # save the image cv2.imwrite (input_img + ". Png", img_new) ./9.jpg ') # Save to another location # Display pictures, call opencv Show # cv2.imshow ("img_new", img_new) # cv2.waitKey (0) # cv2.destroyAllWindows () # Display pictures, call matplotlib .pyplot Show plt.subplot (121), plt.imshow (img_convert (img), cmap = 'gray'), plt.title ('IMG') plt.subplot (122), plt.imshow (img_convert (img_new), plt.title ('IMG_NEW') plt.show () # cv2 Image conversion with matplotlib, cv2 is bgr format, matplotlib is rgb format def img_convert (cv2_img): # grayscale image directly returns if len (cv2_img.shape) == 2: return cv2_img # 3 channel BGR picture elif len (cv2_img.shape) == 3 and cv2_img.shape  == 3: b, g, r = cv2.split (cv2_img ) return cv2.merge ((r, g, b)) # 4 Channel BGR picture elif len (cv2_img.shape) == 3 and cv2_img.shape  == 4: b, g, r, a = cv2 .split (cv2_img) return cv2.merge ((r, g, b, a)) # unknown image format else: return cv2_img # main function if __name__ == "__main__": img_path = gb.glob ("img / *") for path in img_path: print path img_deal (path) 3, Effect 4, Reference Help Documents OpenCV complete sets of information is relatively small, you also need to see help documentation >>> >>> from matplotlib import pyplot Backend TkAgg is interactive backend. function imshow in module matplotlib.pyplot: imshow (X, cmap = None, norm = None, aspect = None, interpolation = None, alpha = None, vmin = None, vmax = None, origin = None, extent = None, None, filradorm = 1, filterrad = 4.0, imlim = None, resample = None, url = None, hold = None, ** kwargs) call signature :: imshow (X, cmap = None, interpolation = None, alpha = None, vmin = None, vmax = None, origin = None, extent = None, ** kwargs) Display the image in * X * to current axes. * X * may be a float array, a uint8 array or a PIL image. If * X * is an array, * X * can have the following shapes: * MxN - luminance (grayscale, float array only) * MxNx3 - RGB (float or uint8 array) * MxNx4 - RGBA (float or uint8 array) The value for each component of MxNx3 and MxNx4 float arrays should be in the range 0.0 to 1.0; MxN float arrays may be normalized. An: class: `matplotlib.image.AxesImage` instance is returned.