尝试写了一个去除图片周围像素填充的函数，大概测试了一下，功能没啥问题，可以去除四边任意的像素值的填充，当然会先检查一下的，对于上下或者左右的边框去除，有两种模式，一种是可以去除两边不同像素值的，一个是去除限定相同像素值的。
不过测试的图片是四周都是[15,15,15]的填充
from PIL import Image, PngImagePlugin
import numpy as np
def crop_img_idex(mask, begin, stop, order = True):
if order:
sequen = 1
else:
sequen = -1
id = 0 # 这里不先赋值 就会报错local variable 'id' referenced before assignment
for id in range(begin, stop, sequen):
if mask[id] > 0:
break
return id
def whether_to_rm(mask, bottom = False):
to_rm = True
if not bottom:
if mask[1] > 0:
to_rm = False
print("无边框可以移除")
elif bottom:
if mask[-1] >0:
to_rm =False
print("无边框可以移除")
return to_rm
def remove_one_edge(Img, imarray, top = True, bottom = False, left = False, right = False, diff = False):
# top 0 h-1 bottom h-1 0 left 0 w-1
w = Img.width # 图片的宽
h = Img.height # 图片的高
w_begin, h_begin, w_new, h_new = 0, 0, w, h
if top or left:
judge = imarray[0, 0]
elif bottom == 1 or right:
judge = imarray[h - 1, w - 1]
mask = np.sum(imarray != judge, axis=-1) > 0
if top or bottom:
mask1 = np.sum(mask, axis=-1)
if not whether_to_rm(mask1, bottom=bottom):
return Img
if top:
h_begin = crop_img_idex(mask1, 0, h)
h_begin = h_begin + 1
print(f"将要剪裁上边界至像素{h_begin}")
elif bottom:
h_new = crop_img_idex(mask1, h-1, 0, order=False)
h_new = h_new - 1
print(f"将要剪裁下边界至像素{h_new}")
if left or right:
mask2 = np.sum(mask, axis=-2)
if not whether_to_rm(mask2,bottom=right):
return Img
if left:
w_begin = crop_img_idex(mask2, 0, w)
w_begin = w_begin + 1
print(f"将要剪裁左边界至像素{w_begin}")
elif right:
w_new = crop_img_idex(mask2, w-1, 0,order=False)
w_new = w_new - 1
print(f"将要剪裁右边界至像素{w_new}")
Img = Img.crop((w_begin, h_begin, w_new, h_new))
return Img
def rm_edges(Img,imarray,edge ):
w = Img.width # 图片的宽
h = Img.height # 图片的高
w_begin, h_begin, w_new, h_new = 0, 0, w, h
if edge.all == np.array([0,1,0,1]).all:
judge = imarray[h - 1, w - 1]
else:
judge = imarray[0, 0]
mask = np.sum(imarray != judge, axis=-1) > 0
maskud = np.sum(mask, axis=-1)
masklr = np.sum(mask, axis=-2)
if edge[0]==1:
h_begin = crop_img_idex(maskud, 0, h)
h_begin = h_begin + 1
print(f"将要剪裁上边界至像素{h_begin}")
if edge[1]==1:
h_new = crop_img_idex(maskud, h - 1, 0, order=False)
h_new = h_new - 1
print(f"将要剪裁下边界至像素{h_new}")
if edge[2]==1:
w_begin = crop_img_idex(masklr, 0, w)
w_begin = w_begin + 1
print(f"将要剪裁左边界至像素{w_begin}")
if edge[3]==1:
w_new = crop_img_idex(masklr, w - 1, 0, order=False)
w_new = w_new - 1
print(f"将要剪裁右边界至像素{w_new}")
Img = Img.crop((w_begin, h_begin, w_new, h_new))
return Img
def remove_edge(img, edge = None, style = "diff"):
"""
img: 传入图片，格式为PIL格式，np格式，或者是路径
edge: 为一个list[x,x,x,x] x为0或者1，如果为1则代表去除这条边的边框，顺序分别为上下左右，同时为1的两条边的边界可以是一样的也可以是不一样的，
默认选项为diff
style:["same", "diff"],same 表示去除的几个边的框框的像素值需要是一样的，diff表示只要框的一条边上像素值是一样的就可以去除
当edge没传参的时候,默认按照上下左右顺序，尽可能去掉大宽度的边框
"""
if type(img) == str:
img = Image.open(img)
imarray = np.array(img)
elif type(img) is np.ndarray:
imarray = img
img = Image.fromarray(imarray)
elif type(img) is PngImagePlugin.PngImageFile:
imarray = np.array(img)
else:
print("对象既不是地址，也不是数组，PngImageFile")
return 0
#print(type(edge))
if type(edge) == list:
edge = np.array(edge)
print( (type(edge) is np.ndarray and (edge.ndim != 1 or edge.shape != (4,))) )
if (type(edge) is None) and (type(edge) is not np.ndarray or (type(edge) is np.ndarray and (edge.ndim != 1 or edge.shape != (4,)))):
print("edge 传入参数错误！")
return img
if edge is not None and sum(edge) != sum(edge == 1):
# 修正edge的数值
for i in [0, 1, 2, 3]:
if edge[i]<0:
edge[i] = 0
elif edge[i]>0:
edge[i] = 1
if edge is not None and sum(edge) == 0:
print("无需修改")
return img
w = img.width # 图片的宽
h = img.height # 图片的高
if edge is not None and sum(edge) == 1:
# 这个情况下 无论是diff还是same都是一样的
img = remove_one_edge(img, imarray, edge[0]==1, edge[1]==1, edge[2]==1, edge[3]==1)
print(1)
elif edge is not None and sum(edge) == 2:
edge1 = edge.reshape(2, 2)
if np.sum((np.sum(edge1, axis=-1) == 2)):
# 这里是对边 对边是区分diff和 same的
print("这里是对边")
if style == "diff":
img = remove_one_edge(img, imarray, edge[0]==1, False, edge[2]==1, False)
imarray = np.array(img)
img = remove_one_edge(img, imarray, False, edge[1] == 1, False, edge[3] == 1)
elif style == "same":
img = rm_edges(img, imarray, edge)
elif edge is not None and np.sum((np.sum(edge, axis=-1) == 2)):
#这里是临边 临边其实不区分
print("这里是临边")
img = rm_edges(img, imarray, edge)
elif edge is not None and sum(edge) >= 2:
img = rm_edges(img, imarray, edge)
elif edge == None:
img = remove_one_edge(img, imarray, True, False, True, False)
imarray = np.array(img)
img = remove_one_edge(img, imarray, False, True, False, True)
return img
path = "E:\\docu\\QQ\\file\\frame000.png"
img = Image.open(path)
#imarray = np.array(img)
# type(imarray) imarray.shape imarray.ndim
img = remove_edge(img, [1,0,1,0])
print(img.width, img.height)
img.show()
