Any method to remove stroke fill color in a jpg image?

[nissansz]

[radarhere]

Pillow can filter out specific colours and modify specific pixels, but as far as determining automatically which pixels are part of the stroke,

#7376 (comment)

Pillow doesn't have any functionality for text detection.

#4572 (comment)

We don't currently ship any techniques for edge detection, so I'm not sure that this is the appropriate library for that. OpenCV or SciPy would probably be better choices.

[nissansz]

I realized that this effect is slightly different from stroke color. It is more heavy shading. Any method to draw such effect in pillow with text?

…

--------------------------------------------------------------------------------

------------------ 原始邮件 ------------------ 发件人: Andrew Murray ***@***.***> 发送时间: 2024-06-12 06:22:33 收件人:python-pillow/Pillow ***@***.***> 抄送:nissanjp ***@***.***>,Author ***@***.***> 主题: Re: [python-pillow/Pillow] Any method to remove stroke fill color ina jpg image? (Discussion #8133) Pillow can filter out specific colours and modify specific pixels, but as far as determining automatically which pixels are part of the stroke, #7376 (comment) Pillow doesn't have any functionality for text detection. #4572 (comment) We don't currently ship any techniques for edge detection, so I'm not sure that this is the appropriate library for that. OpenCV or SciPy would probably be better choices. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[nissansz]

Do you mean crop is faster than resize to get a target size image for drawing?

original jpg background image size is 1280x720, how to crop a text-line length subimage for draw?
Is below method fine?

def random_crop_text_line(image_path, text_length):
    # Open the image
    img = Image.open(image_path)
    
    # Get the width and height of the image
    width, height = img.size
    
    # Define the maximum x-coordinate to crop from
    max_x = width - text_length
    
    # Randomly select the starting x-coordinate for cropping
    start_x = random.randint(0, max_x)
    
    # Crop the subimage with text-line length
    subimg = img.crop((start_x, 0, start_x + text_length, height))
    
    return subimg

[radarhere]

I meant that if your goal is only to blur the text, and not the background, then blurring a smaller image would be faster than blurring a larger image.

from PIL import Image, ImageDraw, ImageFilter, ImageFont
import time
im = Image.new("RGB", (1000, 1000), "#f00")
font = ImageFont.load_default(48)

# Drawing the text and then blurring the whole image is slower
start = time.time()
d = ImageDraw.Draw(im)
d.text((10, 10), "text", fill="#000", font=font)
im = im.filter(ImageFilter.BLUR)
print(time.time() - start)

im = Image.new("RGB", (1000, 1000), "#f00")

# Blurring just the text is faster
start = time.time()
left, top, right, bottom = font.getbbox("text")
text_im = Image.new("RGBA", (right + 20, bottom + 20))
d = ImageDraw.Draw(text_im)
d.text((10, 10), "text", fill="#000", font=font)
text_im = text_im.filter(ImageFilter.BLUR)
text_im.save("out.png")
im.paste(text_im, mask=text_im)
print(time.time() - start)

[nissansz]

So, I need 3 layers?

1. crop a colorful background.
2. blur a text layer
3. paste blue layer on background
4. draw clear text on step 3?

[radarhere]

I don't know what you are trying to achieve or what steps you need to get there - this is the first time you've talked about anything blue in this conversation.

I'm just saying that blurring only the text would be faster than blurring a large image.

[nissansz]

Ok. Thank you.