COMPRESSING >1GB Images with LOSSLESS SOFTWSRE. Please help.

[JeremyVN]

Hi guys,

Do you know how to compress >1GB images without losing quality?

I have a mosaic image and want to share with my colleagues through a group chat.

Instead of uploading to dropbox, or google drive. I just need to upload in group chat and my friends can view it immediately.

Thank you so much.

[johns0]

What format is the image?Any conversion will reduce quality.

[jagabo]

If your images are already compressed (for example JPEG images) you cannot compress them losslessly any more. If they are uncompressed (for example BMP images) you can compress them with a lossless container/codec, like PNG. Depending on the nature of the images they may compress to about 1/2 to 1/4 the source size (real world photographs) or much more (things like computer generated bar charts).

[Cornucopia]

Yes, most forms of lossy compression ALREADY have a component/algorithm of lossless compression built into them, with the lossy part (the part that throws out visually less perceptible material) being the part that FURTHER reduces the size. So, trying to "losslessly compress" that lossy file very often results in files that are LARGER than their lossy source.

Scott

[JeremyVN]

Hi johns0, jagabo and Cornucopia,

It is a png file. The link is attached below. I hope I can compress this file more and make it viewable on the sharing group chat instead of downloading for view.

https://www.dropbox.com/s/4ipm6fnfi0wfx2s/07302022.png?dl=0

What format is the image?Any conversion will reduce quality.

[johns0]

Psp 2018 opened the file but kept getting erors trying to save as jpp,saved as bmp and the file went to 1.58 gb in size,so i don't think it will reduce in size,just too many pictures embedded.

[jagabo]

Your PNG image is already losslessly compressed. You're not likely to find another image format with lossless compression that can give a smaller file.

[johns0]

Best thing to do is split the image,i found that the size was 1920x297000,so it works out to be 275 images of1920x1080.

[JeremyVN]

[Attachment 65733 - Click to enlarge]Hi johns0, jagabo

There were 9000 images embedded in this image. This file was for yesterday.

Today I have 12000 keyframes.

I used you guys' advices. I divided it into 3 parts. But the file is more than 400MB and is bigger than my expectation.

Do you think I can use pngcrush to compress it more? I tried based on instructions from youtube. https://www.youtube.com/watch?v=ORSwlbce5rw&t=82s and https://www.youtube.com/watch?v=V0UCNuENdi8

But it does not work. Image is attached below.

Psp 2018 opened the file but kept getting erors trying to save as jpp,saved as bmp and the file went to 1.58 gb in size,so i don't think it will reduce in size,just too many pictures embedded.

[poisondeathray]

pngcrush or optipng might reduce typical png filesize in the order of 0-3% on average

lossless webp usually reduces typical png filesize in the order of 10-30%. But the maximum dimension for width or height for webp is 16383

[johns0]

If using lossless webp the original png would have to be split 18.2 times height wise.

[johns0]

Also did some calculation and if you used cuts without any shrinking in size and saved in jpg format the file size would come to about 300mB for all the cuts.

[poisondeathray]

I cropped a top 1920x1080 test patch from the png using imagemagick

Here are some lossless RGB compression image benchmarks

png 2,682,915 bytes
pngcrush 2,655,999 bytes
optipng o7 (very slow brute force setting) 2,655,939 bytes
webp lossless rgb 2,089,680 bytes
webp2 lossless rgb 1,947,363 bytes (webp2 is experimental, not widely supported yet)
avif lossless rgb 2,860,688 bytes

webp is widely supported by modern browsers, but the decoding speed is slower than png (high latency image format), and of course the dimension limitation

Do you really need images that large in a chat setting ? Do you really need "lossless" format in a chat setting ? (that example was from a video, but the images are likely downscaled, so it's hardly "lossless") . Also, it's still being downloaded in the chat to be viewed. Uploading large images to a chat seems like bad "netiquette" ? I'd kick someone out of the chat for doing that

[jagabo]

I ran the 973 MiB1920x29700 png file through pngcrush and got a 672 MiB PNG file, a 30 percent reduction. That's better than I expected. It took something like 30 minutes though.

But you should probably stick to more common image sizes like 3840x2160 or 7680x4320. And why do you need lossless encoding for thumbnails? JPEG at 75 percent quality isn't bad and each 3840x2160 image would turn out around maybe 1.3 MB.


[Attachment 65734 - Click to enlarge]

[JeremyVN]

Hi jagabo, poisondeathway, johns0,

Firstly, Thank you so much for spending your time to answer my question and to benchmark the image for me.

You guys are teaching and recommending me about other softwares I haven't known before.

The reasons I make the thumbnail with so many images embedded inside because

For videoeditor and cammeraman, I can learn how other stations edit their videos. How they put their sequences. Whether or not they used a lot of close-up shots. Comparing between ad shot and news shot...etc. Thumbnail gives me an overview and a summary for analyzing the news footage. Watching 5 hour news video is a little challenge and a motivation killer.

This thumbnail has 9000 keyframes from 4-hour news footage. If I stick with a common sizes like 3840x2160 or 7680x4320. There will be too many images; As a result, I want a image with all keyframes embedded inside, but maintaining the quality. When I want to read the news only, I just go right to there and zoom-in.

[Cornucopia]

1st - Converting from PNG to an HEIF picture format file OUGHT to compress it much further, even in lossless mode, although you should SERIOUSLY reconsider your expectations of the downsides of (mild) lossy encoding. Note however, the HEIF format will be much more taxing on your computer to decode - especially at those and similar dimensions - compared to PNG.

2nd - You need to adjust your expectations about what is necessary in these composites. Different titles/captures will have differing amounts of keyframes which will vary greatly the filesize, especially if the runtimes are different as well. It would be much more consistent to constrain the files to certain runtimes (e.g. 2 hours) and have a set allowed (or allowed max) # of keyframes. And keyframes might not even be necessary - are you expecting to view only "poster" frames of major scene changes (edits)? A frame differencing comparison method might winnow out more to give only the more "iconic" frames in the sequence, leaving you less to have to sift through.

3rd - You mention camera folks and editing folks, but they are separate tasks done (usually) by separate people with separate skillsets. I would suggest focusing on one of these things and then once you have gotten a further understanding of that topic, only then focusing on the other. This will affect your criteria for filebuilding.

4th - You need to re-adjust your expectations of what these file formats (and your computer) are reasonably capable of. Large (and odd) canvas sizes like these are very taxing on computer resources, and that ends up slowing down load times, etc, which is counter to the time-saving workflow you think your are going to achieve.

5th - Ultimately, if what you are trying to do is understand how to shoot and edit, I'm going to have to break it to you with a phrase given fairly often in the industry: "the only way to really get good at knowing how to edit and shoot properly is by actually doing the shooting and the editing." An many (most?) of those pros got their start with a college degree in media studies, where they learned the fundamentals of all this, which is NOT insignificant and takes quite a bit to soak in and get straight. For example, you might be finding some statistical significances that might lead you to think "this editor decided to switch to this shot here" but you likely won't get from stats WHY they made that decision.

Scott