ocr.tar.gz_OCR_made_ocr python_python ocr

import Image import time import math import sys sys.setrecursionlimit(20000) class OCR: def __init__(self, image_fn): self.image_fn = image_fn self.im = Image.open(image_fn) self.pixels = self.im.load() self.debug_im = Image.open(image_fn) self.debug_pixels = self.debug_im.load() def start(self): self.done = [] #self.get_borders() run = self.run_vector(FindBorder, (0,0), (1,1), True) self.get_borders(run.cpixel) self.im.save("/tmp/aa.png") self.debug_im.save("/tmp/bb.png") def run_vector(self, runner, *args, **kwargs): run = runner(self.pixels, self.debug_pixels, *args, **kwargs) return run def similar_pixels_by_vector(self, run_vector, max_diff=30, max_distance=10): pixel = run_vector.cpixel def similar_pixels(self, pixel, max_diff=30): similars = {} color = self.pixels[pixel] for pixel2 in self.around_pixels(pixel): try: color2 = self.pixels[pixel2] except IndexError: print pixel2, self.im.size time.sleep(2) continue # the difference for each RGB values in the two pixels diff = map(lambda a,b: abs(a-b), color, color2) average_diff = (diff[0]+diff[1]+diff[2])/3 if average_diff < max_diff: similars[pixel2] = average_diff # order by most similar pixels first similars_sorted = sorted(similars, key=similars.__getitem__) return similars_sorted def around_pixels(self, pixel): around = [] for i in range(8): # pixels start in the up left corner and go on clockwise if i == 0: pixel2 = (pixel[0]-1, pixel[1]-1) if i == 1: pixel2 = (pixel[0], pixel[1]-1) if i == 2: pixel2 = (pixel[0]+1, pixel[1]-1) if i == 3: pixel2 = (pixel[0]+1, pixel[1]) if i == 4: pixel2 = (pixel[0]+1, pixel[1]+1) if i == 5: pixel2 = (pixel[0], pixel[1]+1) if i == 6: pixel2 = (pixel[0]-1, pixel[1]+1) if i == 7: pixel2 = (pixel[0]-1, pixel[1]) if pixel2[0] < 0 or pixel2[1] < 0 or \ pixel2[0] >= self.im.size[0] or pixel2[1] >= self.im.size[1]: continue else: around.append(pixel2) return around def get_borders(self, pixel): #run = self.run_vector(FindBorder, (self.im.size[0]-1,self.im.size[1]-1),(-1,-1), True) #run = self.run_vector(FindBorder, (self.im.size[0]-1,0),(-1,1), True) #run = self.run_vector(FindBorder, (0,self.im.size[1]-1),(1,-1), True) self.debug_pixels[pixel] = (255,0,0) self.done.append(pixel) sim = self.similar_pixels(pixel, 40) if not len(self.done)%1000: print len(sim), len(self.done) for x in sim: if x not in self.done: self.get_borders(x) def fg_color(self): pass class RunVector: def __init__(self, pixels, debug_pixels, start, vector, run_now=False, callback=None): self.pixels = pixels self.debug_pixels = debug_pixels self.vector = vector self.cpos = start # cpos->current position self.cpixel = start # cpixel->current pixel self.trace = [start] if callback: self.callback = callback if run_now: self.run() def debug(self, msg, error=False): out_msg = "" if type(msg) in [list, tuple]: for x in msg: if type(x) != str: x = repr(x) out_msg += x+" " elif type(msg) != str: out_msg = repr(msg) else: out_msg = msg if error: error = "==ERROR== " else: error = "" print "%s<%s>: %s" % (error, self.__class__.__name__, out_msg) def run(self): if max([abs(x) for x in self.vector]) == abs(self.vector[0]): maxv = "x" diff = abs(float(self.vector[1]) / self.vector[0]) else: maxv = "y" diff = abs(float(self.vector[0]) / self.vector[1]) while 1: if len(self.trace) > 1: if maxv == "x": self.cpos = self.cpos[0]+math.copysign(1, self.vector[0]),\ self.cpos[1]+math.copysign(diff,self.vector[1]) elif maxv == "y": self.cpos=self.cpos[0]+math.copysign(diff,self.vector[0]),\ self.cpos[1]+math.copysign(1, self.vector[1]) self.cpixel = int(self.cpos[0]), int(self.cpos[1]) if self.callback(): return else: self.trace.append(self.cpixel) def trace_colors(self): return [self.pixels[x] for x in self.trace] class FindBorder(RunVector): def callback(self): #self.debug((self.cpixel, self.pixels[self.cpixel])) if self.pixels[self.cpixel] not in self.trace_colors(): self.debug(("Found a border", self.cpixel, self.pixels[self.cpixel])) return 1 self.debug_pixels[self.cpixel] = (255,0,0) a = OCR("getimage.png") a.start()