supplierdeeply

Found over the Internet: Time Domain Continuous Imaging with CHDK.

  • 3 Replies
  • 523 Views
Advertisements

Hi! Just found this one today. Not being sure if you are aware of the paper I thought it's a good idea to share the stuff with you.

A few words from the abstract:
Quote
Time Domain Continuous Imaging (TDCI) is a new model for photography that allows exposure timing to be freely manipulated after capture. This is done by creating, and operating on, a continuous waveform representation of how the value of each pixel changes over time.

http://www.ingentaconnect.com/contentone/ist/ei/2017/00002017/00000015/art00011

They implement the thing using the live view, but would be really AMAZING to implement that from the actual image buffer.
if (2*b || !2*b) {
    cout<<question
}

Compile error: poor Yorick

Re: Found over the Internet: Time Domain Continuous Imaging with CHDK.
« Reply #1 on: 17 / September / 2017, 15:56:23 »
Thanks, interesting, I'll have a look into it. Any idea where the scripts they are using are documented?

Cheers, Lee

*

Online reyalp

  • ******
  • 11084
Re: Found over the Internet: Time Domain Continuous Imaging with CHDK.
« Reply #2 on: 17 / September / 2017, 16:08:36 »
This is the work of @ProfHankD who occasionally posts here and his students.

Quote
Any idea where the scripts they are using are documented?
They are using custom C code, described in the paper and apparently available from http://aggregate.org/DIT/TIK/

edit:
Note the source is in a ~ 300mb (1gb uncompressed) tar gz file of a full build tree. I've taken the liberty of converting this to a patch against r4805 trunk here.

I have not attempted to run this or use the tik tools.
« Last Edit: 17 / September / 2017, 17:17:50 by reyalp »
Don't forget what the H stands for.

Re: Found over the Internet: Time Domain Continuous Imaging with CHDK.
« Reply #3 on: 18 / September / 2017, 11:44:17 »
The CHDK TDCI in-camera stuff is still very fragile and has portability issues, which is why we haven't widely distributed it.


The problem is where we can get enough memory for working space. Currently, it uses what would be the still image capture frame buffer as buffer space for building-up the TDCI representation of the live view data. However, even that fails on some cameras (e.g., the SX530HS) because the image buffer gets overwritten despite the fact that we're not capturing any still image. BTW, working from the still image buffer data directly is something we've also implemented, but it doesn't produce a dramatically faster capture rate than dumb saving of the still buffer.


Anyway, we'll post better versions once they're a bit more stable....





 

Related Topics