Some black pictures on html

[Yichen-fqyd]

Hi, when I was converting pdf file to html file, sometimes there appears a total black image shown on html, while original pdf didn't show,
What is the reason for this issue, and how to possibly solve it?
is there a way to get information from block==2 get("dict") to know that it is a totally black image?
{'bbox': (386.1600036621094, 82.44001007080078, 491.6400146484375, 103.32000732421875), 'bpc': 8, 'colorspace': 3, 'ext': 'png', 'height': 57, 'image': b'\x89PNG\r\n\x1a\n\x00\x00\x00\rIHDR\x00\x00\x01%...\x00\x00IEND\xaeB`\x82', 'number': 10, 'type': 1, 'width': 293, 'xres': 96, 'yres': 96}

[JorjMcKie]

This a known bug - see #657. The image has an alpha channel. You can find this out by inspecting the list page.get_images(). For each image on the page, there is an item like (xref, smask, ...) in this list. If smask is > 0, then this image may show this unpleasant appearance in HTML output.

[JorjMcKie]

The coming new version 1.18.11 will contain a new Page method get_image_info(). It reports the same content as the get_text("dict") method for images. But it will not return text and not the binary image content - only the respective images sizes.
In addition, there will be thea new dictionary item transform, which will contain information about what happened to the original image to make it fit into that page's bbox - IAW scaling or rotation can then be detected in detail.

[Yichen-fqyd]

Thank you for your detailed explanation, now I am clear.

Looking forward the new version 1.18.11, I am pretty sure this function get_image_info() will be very helpful!!!

[Yichen-fqyd]

As a consequence of the post before: your PDF page must contain 2 inline images ...

Hi,
Follow up for the inline images, if the inline images only present in the image blocks, how can I get information whether the image has mask or alpha channel,

Thank you

[JorjMcKie]

There currently is no way to determine whether an inline image has an alpha channel - and consequently there is no way to access the alpha information.
For other images, this can be found out by inspecting the items of list page.get_images(). You can use page.get_image_info(xref=True) which is a list of image meta info dictionaries displayed by the page. Those dictionaries contain an "xref" key, which points to a non-inline image if >0.

[JorjMcKie]

Additional note to my previous post:
I have never checked this out: MuPDF detects inline images and returns the corresponding binary as part of the image block in page.get_text("dict"). There is a chance, that this logic is smart enough to include any existing alpha channel. Can be checked by making a pixmap of that bytes object and look at pix.alpha.