汉字字库读取

#!/usr/bin/python

#encoding: utf-8

import numpy as np

RECT_HEIGHT = 16

RECT_WIDTH = 16

BYTE_COUNT_PER_ROW = int(RECT_WIDTH / 8)

BYTE_COUNT_PER_FONT = BYTE_COUNT_PER_ROW * RECT_HEIGHT

KEYS = [0x80, 0x40, 0x20, 0x10, 0x08, 0x04, 0x02, 0x01]

class FontRender(object):

def __init__(self, font_file,

rect_height=RECT_HEIGHT, rect_width=RECT_WIDTH, byte_count_per_row=BYTE_COUNT_PER_ROW):

self.font_file = font_file

self.rect_height = rect_height

self.rect_width = rect_width

self.byte_count_per_row = byte_count_per_row

with open(font_file, "rb") as f:

self.data = f.read()

self.fonts=[]

for idx in range(0,len(self.data),BYTE_COUNT_PER_FONT):

font_rect=self.data[idx:idx+BYTE_COUNT_PER_FONT]

font=self.font2np(font_rect)

self.fonts.append(font)

 

self.__init_rect_list__()

def __init_rect_list__(self):

self.rect_list = [] * RECT_HEIGHT

for i in range(RECT_HEIGHT):

self.rect_list.append([] * RECT_WIDTH)

def get_font_area_index(self, txt, encoding='utf-8'):

# if not isinstance(txt, unicode):

# txt = txt.decode(encoding)

gb2312 = txt.encode('gb2312')

# hex_str = binascii.b2a_hex(gb2312)

# area = eval('0x' + hex_str[:2]) - 0xA0

# index = eval('0x' + hex_str[2:]) - 0xA0

area = gb2312[0] - 0xA0

index = gb2312[1] - 0xA0

return area, index

def get_font_rect(self, area, index):

offset =int( (94 * (area-1) + (index-1)) * BYTE_COUNT_PER_FONT )

btxt = self.data[offset:offset+32]

# with open(self.font_file, "rb") as f:

# f.seek(offset)

# btxt = f.read(BYTE_COUNT_PER_FONT)

return btxt

def convert_font_rect(self, font_rect, ft=1, ff=0):

for k in range(RECT_HEIGHT):

row_list = self.rect_list[k]

for j in range(self.byte_count_per_row):

idx=k * self.byte_count_per_row + j

asc = font_rect[idx]

for i in range(8):

flag = asc & KEYS[i]

row_list.append(flag and ft or ff)

def font2np(self,font_rect):

data=np.zeros((RECT_HEIGHT,RECT_HEIGHT),dtype=np.float32)

for k in range(RECT_HEIGHT):

for j in range(self.byte_count_per_row):

idx=k * self.byte_count_per_row + j

asc = font_rect[idx]

for i in range(8):

flag = asc & KEYS[i]

if flag:

data[k,j*8+i]= 1

return data

def render_font_rect(self, rect_list=None):

if not rect_list:

rect_list = self.rect_list

for row in rect_list:

for i in row:

if i:

print ('■'),

else:

print ('○'),

print

def convert(self, text, encoding='utf-8'):

# if not isinstance(text, unicodedata):

# text = text.decode(encoding)

data=[]

for t in text:

#area, index = self.get_font_area_index(t)

#font_rect = self.get_font_rect(area, index)

font=self.getFont(t)

data.append(font)

return data

def getFont(self,hz):

"""hz:汉字"""

area, index = self.get_font_area_index(hz)

font_rect = self.get_font_rect(area, index)

font=self.font2np(font_rect)

return font

def getAt(self,i):

# idx = i*BYTE_COUNT_PER_FONT

# font_rect=self.data[idx:idx+BYTE_COUNT_PER_FONT]

# font=self.font2np(font_rect)

return self.fonts[i]

def getall(self):

data=[]

for i in range(0,len(self.data)//BYTE_COUNT_PER_FONT+1):

font = self.getAt(i)

data.append(font)

return data

def get_rect_info(self):

return self.rect_list

import torch.utils.data as data

from typing import Any, Callable, List, Optional, Tuple

class HzDataset(data.Dataset):

def __init__(self) -> None:

super().__init__()

self.fr = FontRender('./HZK16')

self.cnt = len(self.fr.data)//BYTE_COUNT_PER_FONT

 

def __getitem__(self, index: int) -> Any:

"""

区位码＝区号＋位号［1-94,1-94］ 国标码＝区位码＋32［ (1-94)+32， (1-94)+32］ 机内码＝将国标码的两个字节的第一位置壹。

offset =int( (94 * (area-1) + (index-1)) * BYTE_COUNT_PER_FONT )

area = gb2312[0] - 0xA0

index = gb2312[1] - 0xA0

"""

font=self.fr.getAt(index)

area=index//94+1

idx=index%94+1 #txt.encode('gb2312')

a=bytes([area+0xA0,idx+0xA0])

#label = bytes.decode(a,'gb2312')

return font[np.newaxis,:,:],a #label

def __len__(self) -> int:

return self.cnt

def test():

hz='./pallium/HZK16'

with open(hz, "rb") as f:

btxt = f.read(BYTE_COUNT_PER_FONT)

 

if '__main__' == __name__:

import cv2

text = u'同创伟业'

fr = FontRender('./HZK16')

data = fr.convert(text)

cv2.imshow("aaa", np.hstack(data))

cv2.waitKey(0)

data = fr.getall()

cv2.imshow("aaa", np.hstack(data[0:100]))

cv2.waitKey(0)

下载地址

https://download.csdn.net/download/weixin_38906715/88114756?spm=1001.2014.3001.5503