[Python] OCR in complex image

I need to extract player statistic in a image : enter image description here

I have try pytesseract with preprocessing image : convert to grey scale, resize img with a factor 2 and filter on edge :

with open(TEXT_FILE, "w", encoding="utf-8") as f_out:
for img_file in sorted(FRAMES_DIR.glob("*.png")):
# option : prétraitement léger
#img = Image.open(img_file).convert("L") # niveaux de gris
img = Image.open(img_file)
gray_image = ImageOps.grayscale(img)
# Resize the image to enhance details.
scale_factor = 3
resized_image = gray_image.resize(
(gray_image.width * scale_factor, gray_image.height * scale_factor),
resample=Image.LANCZOS
)
# Apply adaptive thresholding using the `FIND_EDGES` filter.
thresholded_image = gray_image.filter(ImageFilter.FIND_EDGES)

text = pytesseract.image_to_string(
thresholded_image,
lang="eng", # français (ajoute "eng" si besoin)
#config="--psm 6" # adapté pour du texte en ligne / interface
config = tessdata_config
).strip()

if text:
f_out.write(f"===== {img_file.name} =====\n")
f_out.write(text + "\n\n")

print(f"OCR terminé. Résultat dans : {TEXT_FILE}")


but the result was not good.

Can you give me some advise to this task or an other lib to extract text

regards

Continue reading...