# coding: utf-8
import os
import cv2
import warnings
import numpy as np
from .drawing import cv2WHITE
from ..utils.generic_utils import filenaming
from ..utils._colorings import toBLUE
[docs]def cv2paste(bg_img, fg_img, points=(0,0), inplace=False):
"""Pastes ``fg_image`` into ``bg_image``
Args:
bg_img (ndarray) : Background Image. shape=(H,W,ch)
fg_img (ndarray) : Background Image. shape=(H,W,ch)
points (tuple) : Coordinates to paste. (x,y)
inplace (bool) : Whether to transform input ( ``bg_img`` ) using no auxiliary data structure.
Returns:
bg_img (ndarray) : pasted image.
Examples:
>>> import cv2
>>> from pycharmers.opencv import SAMPLE_LENA_IMG, cv2read_mpl, cv2plot, cv2paste
>>> bg_img = cv2read_mpl(SAMPLE_LENA_IMG)
>>> fg_img = cv2.resize(bg_img, dsize=(256,256))
>>> ax = cv2plot(cv2paste(bg_img, fg_img, points=(128,128)))
"""
if not inplace:
bg_img = bg_img.copy()
x,y = points
bg_h, bg_w, _ = bg_img.shape
fg_h, fg_w, _ = fg_img.shape
if ((-fg_w < x < bg_w) and (-fg_h < y < bg_h)):
if not inplace:
bg_img = bg_img.copy()
bg_img[max(0,y):min(y+fg_h, bg_h), max(0,x):min(x+fg_w, bg_w), :] = fg_img[max(0,0-y):bg_h-y, max(0,0-x):bg_w-x, :]
return bg_img
[docs]def vconcat_resize_min(*images, interpolation=cv2.INTER_CUBIC):
"""Concat vertically while resizing to the smallest width.
Args:
images (np.ndarray) : OpenCV images
interpolation (int) : interpolation method, see `OpenCV Documentations #InterpolationFlags <https://docs.opencv.org/master/da/d54/group__imgproc__transform.html#ga5bb5a1fea74ea38e1a5445ca803ff121>`_
Examples:
>>> import cv2
>>> from pycharmers.opencv import vconcat_resize_min, cv2plot
>>> images = [cv2.imread(path) for path in os.listdir("images")]
>>> vconcat_img = vconcat_resize_min(*images)
>>> ax = cv2plot(vconcat_img)
"""
w_min = min(img.shape[1] for img in images)
return cv2.vconcat([
cv2.resize(src=img,
dsize=(w_min, int(img.shape[0]*w_min/img.shape[1])),
interpolation=interpolation
) for img in images
])
[docs]def hconcat_resize_min(*images, interpolation=cv2.INTER_CUBIC):
"""Concat horizontally while resizing to the smallest height.
Args:
images (np.ndarray) : OpenCV images
interpolation (int) : interpolation method, see `OpenCV Documentations #InterpolationFlags <https://docs.opencv.org/master/da/d54/group__imgproc__transform.html#ga5bb5a1fea74ea38e1a5445ca803ff121>`_
Examples:
>>> import cv2
>>> from pycharmers.opencv import hconcat_resize_min, cv2plot
>>> images = [cv2.imread(path) for path in os.listdir("images")]
>>> hconcat_img = hconcat_resize_min(*images)
>>> ax = cv2plot(hconcat_img)
"""
h_min = min(img.shape[0] for img in images)
return cv2.hconcat([
cv2.resize(src=img,
dsize=(int(img.shape[1]*h_min/img.shape[0]), h_min),
interpolation=interpolation
) for img in images
])
[docs]def resize_aspect(src, dsize, interpolation=cv2.INTER_AREA):
"""Resize the image while keeping the aspect ratio.
Args:
src (np.ndarray) : Input image.
dsize (tuple) : Output image size ( ``width`` , ``height``)
interpolation (int) : Interpolation method (default= ``cv2.INTER_AREA`` )
Returns:
resized (np.ndarray) : Resized image.
Examples:
>>> import numpy as np
>>> from pycharmers.opencv import resize_aspect
>>> img = np.random.randint(low=0, high=255, size=(1080, 720, 3), dtype=np.uint8)
>>> resized = resize_aspect(src=img, dsize=(300, 300))
>>> resized.shape
(300, 200, 3)
"""
sh, sw = src.shape[:2]
dw, dh = dsize
if sh/sw > dh/dw:
ratio = dh/sh
else:
ratio = dw/sw
dsize = (int(ratio*sw), int(ratio*sh))
resized = cv2.resize(src=src, dsize=dsize, interpolation=interpolation)
return resized
[docs]def transparency(in_path, out_path=None, lower_bgr=cv2WHITE, upper_bgr=cv2WHITE, mode=cv2.RETR_EXTERNAL, method=cv2.CHAIN_APPROX_SIMPLE, thresh=None, check_exist=True):
"""Transparency processing.
Args:
in_path (str) : Path to input image.
out_path (str) : Path to output image.
lower_bgr (tuple/int) : Lower bound of image value to be transparent.
upper_bgr (tuple/int) : Upper bound of image value to be transparent.
mode (int) : Contour retrieval mode used in ``cv2.findContours`` (default = ``cv2.RETR_EXTERNAL`` )
method (int) : ontour approximation method used in ``cv2.findContours`` (default = ``cv2.CHAIN_APPROX_SIMPLE`` )
thresh (int) : Threshold value.
check_exist (bool) : If ``True``, there is a possibility of overwriting the image.
Examples:
>>> from pycharmers.opencv import transparency, SAMPLE_LENA_IMG
>>> transparency(SAMPLE_LENA_IMG)
Saved at /Users/iwasakishuto/.pycharmers/opencv/image/lena_transparency.png
"""
# Naming the output path.
if out_path is None:
root = os.path.splitext(in_path)[0] + "_transparency"
ext = ".png"
else:
root,ext = os.path.splitext(out_path)
if ext==".jpg":
warnings.warn("Since transparent image cannot be created with '.jpg' image, use '.png'.")
ext = ".png"
out_path = root + ext
if check_exist:
out_path = filenaming(out_path)
src = cv2.imread(filename=in_path, flags=cv2.IMREAD_UNCHANGED)
if src.shape[2]==3:
src = np.insert(src, 3, values=[0], axis=2)
if thresh is None:
# Checks if array elements lie between the elements of two other arrays.
binary = 255-cv2.inRange(src=src[:,:,:3], lowerb=np.asarray(lower_bgr), upperb=np.asarray(upper_bgr))
else:
# Thresholding
gray = cv2.imread(filename=in_path, flags=cv2.IMREAD_GRAYSCALE)
binary = cv2.threshold(gray, thresh=thresh, maxval=255, type=cv2.THRESH_BINARY)[1]
contours, _ = cv2.findContours(image=binary, mode=mode, method=method)
mask = np.zeros_like(binary, dtype=np.uint8)
src[:,:,3] = cv2.fillPoly(img=mask, pts=contours, color=255)
if cv2.imwrite(filename=out_path, img=src):
print(f"Saved at {toBLUE(out_path)}")
[docs]def pil2cv(img):
"""Convert ``PIL.Image`` object into ``numpy`` array. (BGR)"""
return cv2.cvtColor(np.asarray(img, dtype=np.uint8), cv2.COLOR_RGBA2BGR)
