Yes check the code from the previous post, It works well with half of the
tesseract psm.
It's just a simple global threshold.
On Wednesday, January 15, 2020 at 5:15:19 AM UTC+1, Richard wrote:
>
> Thank you very much Daniel.
> This is very handy for sure, I can see every step of image modification in
> PyCharm now :)
> I tried to transform image in different orders (crop first, gray, thresh;
> gray first, crop then tresh etc) but still there is no good result.
> Do you have some interesting material that would help to pinpoint what's
> wrong in my preprocess ?
>
> Best regards,
> Richard
>
> Le lundi 13 janvier 2020 17:48:01 UTC+1, Daniel Anton a écrit :
>>
>> Hi Richard,
>>
>> First of all I would recommend you to visualize each step of your process
>> (Threshold, Median ...).
>> I use matplotlib for that propuse you can find an example bellow.
>>
>> The problem is not Tesseract but your preprocess, you shouldn't do a blur
>> to a text that small and for the Threshold you can just use
>> cv2.THRESH_BINARY.
>>
>> import cv2
>> import pytesseract
>>
>> from loguru import logger
>>
>> from matplotlib import pyplot as plt
>>
>>
>> image = cv2.imread("test.png")
>>
>> x = 375
>> y = 0
>> h = 40
>> w = 160
>>
>> # IN GREY
>> image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
>>
>> plt.figure(figsize = (40,40))
>> plt.imshow(image, cmap = "gray")
>> plt.title('image')
>> plt.show()
>>
>>
>> # THRESH
>> image = cv2.threshold(image, 200, 255, cv2.THRESH_BINARY)[1]
>>
>>
>>
>> plt.figure(figsize = (40,40))
>> plt.imshow(image, cmap = "gray")
>> plt.title('image')
>> plt.show()
>>
>> image = image[y:y+h, x:x+w]
>>
>> plt.figure(figsize = (40,40))
>> plt.imshow(image, cmap = "gray")
>> plt.title('image')
>> plt.show()
>>
>> cv2.imwrite("output.png", image)
>>
>> for arg in range(1, 14):
>> try:
>> text = pytesseract.image_to_string(image, config=f"-l eng --oem
>> 1 --psm {arg}")
>> logger.info(f"Parsed text [{text}] with [{arg}]")
>> except Exception as exception:
>> logger.exception(exception)
>>
>>
>>
>>
>> On Monday, January 13, 2020 at 4:18:40 PM UTC+1, Richard wrote:
>>>
>>> Hello,
>>>
>>> I'm starting a software that would analyze eSport matchs, mining
>>> statistics from screenshots (i.e. video).
>>> I've started simple, I'm trying to extract team names.
>>>
>>> I've tried to crop/gray but it's not efficient (i.e. text doesn't match).
>>>
>>> Any tip to improve text recognition ?
>>>
>>> Code sample used:
>>>
>>> import cv2
>>> import pytesseract
>>> from loguru import logger
>>>
>>>
>>> image = cv2.imread("test.png")
>>>
>>> x = 375
>>> y = 0
>>> h = 40
>>> w = 160
>>>
>>> # IN GREY
>>> image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
>>>
>>> # BLUR
>>> image = cv2.medianBlur(image, 3)
>>>
>>> # THRESH
>>> #image = cv2.threshold(image, 0, 255, cv2.THRESH_BINARY |
>>> cv2.THRESH_OTSU)[1]
>>>
>>> #image = image[y:y+h, x:x+w]
>>> cv2.imwrite("output.png", image)
>>>
>>> for arg in range(1, 14):
>>> try:
>>> text = pytesseract.image_to_string(image, config=f"-l eng --oem 1
>>> --psm {arg}")
>>> logger.info(f"Parsed text [{text}] with [{arg}]")
>>> except Exception as exception:
>>> logger.exception(exception)
>>>
>>>
>>>
>>>
--
You received this message because you are subscribed to the Google Groups
"tesseract-ocr" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/tesseract-ocr/840cf11f-bbc5-4fd5-8eb6-43d83c65e163%40googlegroups.com.
