Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Wei Jiang
    @jiangwei221
    Hi, I have a depth image whch is float32, and how can I resize it using PIL? What is the correct way to convert a 2D float32 array to a PIL Image object? Thank you!
    Andrew Murray
    @radarhere

    If you have an image file, I would try -

    from PIL import Image
    im = Image.open('image.tif')
    new_im = im.resize((100, 100))

    If by '2d float32 array' you are referring to a numpy array -

    from PIL import Image
    im = Image.fromarray(x)

    Wei Jiang
    @jiangwei221
    Ah, I see, thanks! I was confused because when I use matplotlib to display the PIL.Image object, it gives me a 4 channel uint8 image. But if I do plt.imshow(np.array(a)), it works perfectly.
    Code:
    depth_map is a 2d float32 np array
    a = Image.fromarray(depth_map, mode='F')
    plt.imshow(a) # gives me a 4 channel uint8 image
    plt.imshow(np.array(a)) # gives the correct visualization
    tkzv
    @tkzv
    Hello. What is the right way to paste images, that start at negative coordinates and use alpha channel? PIL.Image.alpha_composite doesn't accept negative coordinates, and I couldn't figure out how to use PIL.Image.paste with a mask -- it adds the new image under the original one.
    Andrew Murray
    @radarhere
    Here is how I would use paste to do that -
    from PIL import Image
    im_base = Image.open("Tests/images/hopper.jpg")
    im_overlay = Image.open("Tests/images/transparent.png")
    im_base.paste(im_overlay, (-50, -50), im_overlay)
    im_base.save('out.png')
    tkzv
    @tkzv
    I tried that and it didn't look right. But the problem was in the images. Thanks.
    Technerder
    @Technerder
    I'm running into an issue while trying to overlay text over a gif stored in a BytesIO object
                        edited_image_bytes = BytesIO()
                        image = Image.open(BytesIO(await request.read()))
                        frames = []
                        for frame in ImageSequence.Iterator(image):
                            d = ImageDraw.Draw(frame)
                            d.text((10, 100), "Hello World")
                            del d
                            b = BytesIO()
                            frame.save(b, format="GIF")
                            frames.append(Image.open(b))
                        frames[0].save(edited_image_bytes, format='GIF', save_all=True, append_images=frames[1:])
    The file is empty
    Andrew Tolmie
    @DancingQuanta
    Hi, have anyone seeing DLL not found error when piliow is install via conda-forge as dependency of matplotlib or bokeh?
    Andrew Murray
    @radarhere
    I think you’ll be interested in conda-forge/pillow-feedstock#45 and python-pillow/Pillow#4155
    Andrew Tolmie
    @DancingQuanta
    Thanks, I have commented on the recently opened issue conda-forge/pillow-feedstock#66
    Michael Adam Cabatingan
    @Peaceful-Whale
    I am trying to use Pillow with eclipse, but I keep getting this error:
    from PIL import Image
    ImportError: No module named 'PIL'
    I have PIL specified in the external library thing
    Project > Properties > PYDEV - PYTHONPATH
    I have it pointed to the folder I extracted from the download on the Pillow website
    Do I need to point to Pillow - 7.0.0 or PIL in the scr folder of Pillow - 7.0.0
    Michael Adam Cabatingan
    @Peaceful-Whale
    I get the same error if I just delete that reference
    attempting to copy just the PIL folder and use that changes nothing
    I think if I do that and restart I get a different error though
    In the "project references" it shows that it is indeed referencing PIL
    Michael Adam Cabatingan
    @Peaceful-Whale
    I tried to copy all of the files in PIL into my project, but it gives me this error when I try to import them:
    Parent module '' not loaded, cannot perform relative import
    When I stick these in a folder and then call from there
    using from folder import Image
    i get this: ImportError: cannot import name '_imaging'
    Michael Adam Cabatingan
    @Peaceful-Whale
    Image.py is failing on these import lines
    from . import ImageMode, TiffTags, UnidentifiedImageError, __version__, _plugins
    from ._binary import i8, i32le
    from ._util import deferred_error, isPath
    Parent module '' not loaded, cannot perform relative import
    Which i think is just from running Image.py when its inside the package / whatever
    but I thought that might help
    Michael Adam Cabatingan
    @Peaceful-Whale
    I have made a test .py in the folder with everything else and imported it and ran a test func
    so thats working
    it appears to be an error with something in Image.py
    and idk what to do about it. I'm assuming it's becuase its just the PIL folder and not anything else
    i have now imported the entire pillow folder
    have made a test py function, and have imported it
    so I know that works and that I can import files from the PIL folder
    when I try to import Image though i get this
    ImportError: cannot import name '_imaging'
    Michael Adam Cabatingan
    @Peaceful-Whale
    I'm guessing it's something with Image then, and I don't know how to fix it
    Nitin Pasumarthy
    @Nithanaroy

    Is there a way to identify the exact settings for Image.save from original image (to avoid image size on disk inflation)?

    img_path = Path("my_image.jpeg")
    o = Image.open(img_path)
    print(f"Original image size: {img_path.lstat().st_size}")
    ... # perform lossless transformations like crop
    o.save("transformed_img.jpeg", quality=100, subsampling=0)

    This approach leads to an increase in size. I'm not sure how to achieve this.

    Complete code which for now does not do any transformations but opens and saves the image back to disk - https://colab.research.google.com/drive/1pCtAcbZD6YsIzA7vKn7xktF7Hp-leqo7
    Andrew Murray
    @radarhere
    @Nithanaroy considering that even saving and opening a JPG can result in different pixel values - https://github.com/python-pillow/Pillow/issues/3015#issuecomment-368218971 - I wouldn’t think this is possible at the moment with Pillow
    Andrew Murray
    @radarhere
    @Peaceful-Whale https://github.com/python-pillow/Pillow/issues/3115#issuecomment-385713336 is someone with a similar problem in the past, if that’s of any help
    Ak4zh
    @Ak4zh

    Is there a way I can read only selected text froom image that has a colored background?

    I am trying to read price from tradingview charts image.

    For example I want to read only the text with red (0.66133) , green(0.65608) , gray (0.65995) or blue (0.65773) background from this image: https://www.tradingview.com/x/e37LLsNZ Using pytesseract to get text from image
    Andrew Murray
    @radarhere
    To give you a quick answer - if I rephrase your request, you want to crop out parts of an image within a rectangle outline of a certain color. This problem is solved for a different situation in python-pillow/Pillow#4049, using https://pillow.readthedocs.io/en/stable/reference/PixelAccess.html to read individual pixel colors
    Andrew Murray
    @radarhere
    Ok, here is code to process your image and save cropped rectangles with the text. The red, green, gray and blue colors are specified. It will take a minute or two to run. It is searching for a rectangle of one of the colors, of a reasonable size.
    from PIL import Image
    
    colors = ((0,160,0),(33,150,243),(88,88,88),(255,0,0))
    
    with Image.open('tradingview.png') as im:
        px = im.load()
        x, y = 0, 0
        currentColor = None
        processedPixels = []
        i = 0
    
        # Start searching through all the pixels
        for y in range(0, im.height):
            for x in range(0, im.width + 1):
                def getColor(x, y):
                    try:
                        return px[x, y]
                    except IndexError:
                        return
    
                if (x, y) in processedPixels:
                    continue
                color = getColor(x, y)
                if color != currentColor:
                    if currentColor is not None:
                        max_x = x
    
                        # The color line has ended, from min_x to max_x
                        x_range = min_x, max_x
    
                        # The left side of our rectangle must also be a solid color
                        # So search from the top left until we have a solid left side
                        for y2 in range(y+1, im.height+1):
                            for x2 in range(min_x, max_x+1):
                                if getColor(x2, y2) == currentColor:
                                    min_x = x2
                                    break
                            else:
                                # No color in this line, so we have hit the bottom of the rectangle
                                max_y = y2
                                break
    
                        # The same for the right side
                        for y2 in range(y+1, max_y+1):
                            for x2 in range(max_x+1, min_x, -1):
                                if getColor(x2, y2) == currentColor:
                                    max_x = x2
                                    break
    
                        # Check that this rectangle is of a reasonable size
                        if max_x - min_x > 10 and max_y - y > 10:
                            # Record that we've processed these pixels, to save time
                            for y2 in range(y, max_y+1):
                                for x2 in range(x_range[0], x_range[1]+1):
                                    processedPixels.append((x2, y2))
    
                            # Check that the box is not purely one color
                            singleColor = True
                            for y2 in range(y, max_y+1):
                                for x2 in range(min_x, max_x+1):
                                    if getColor(x2, y2) != currentColor:
                                        singleColor = False
                                        break
                                if not singleColor:
                                    break
                            if not singleColor:
                                i += 1
    
                                # Save the cropped image
                                im.crop((min_x, y, max_x, max_y)).save("out_"+str(i)+".png")
                    if color in colors:
                        # Start a line with a relevant color
                        currentColor = color
                        min_x = x
                    else:
                        currentColor = None