Wednesday, May 16, 2012

Idea : Video Enhancement

Recently I have been thinking a lot about Video Enhancement. 

My Thinking Process :

  • Can the Video be enhanced in quality?
  • Can it be added with new information (Pixels) without actually affecting Video?
  • Can a low resolution 3GP video converted to High Quality (HD) Videos ?

Idea : What do I need ?

I was thinking of a Video Converter which enhances the quality of the Videos. As we already know there are Video converters out there which help in converting videos from one format to another with same resolution or lesser.
How about a Video converter which actually takes low resolution videos and delivers high quality (resolution) videos ?
Is this Idea Feasible ?

How can we do that ?

I did do a lot of Survey on the Internet and could not find much regarding this subject.

One implementation goes this way :
It talks about studying the pixel surrounding information in Low quality videos and generating New pixels on the fly based on the surrounding of the pixel and embed it there. This is carried out throughout the Video pixels thus increasing the number of pixels in the Video. Thus enhancing the Video quality and the resolution of the Video.

Some Inputs from the Readers needed :

  • Is this Idea feasible to be implemented with full capacity ?
  • Is there any algorithm for Video enhancements ?
  • Are there any implementations already available ?
  • Any Idea about the Implementation of the Idea ?
DO please leave a comment about what do you think about the Idea .


DataWraith said...

If you have A LOT of time and training material, it should be possible to design a Restricted Boltzmann machine ( to do this.

One idea would be to download random Youtube videos in high and low quality, and use that as a training set for the Boltzmann machine. The machine would learn a probabilistic mapping from low quality video to high quality video, which you could then use to 'guess' missing information in low quality video.

Quality-wise this could work better than simply upscaling the low-quality video, but making it work properly will probably take more effort than the idea is worth in terms of improvement over traditional methods.

Anonymous said...

Lost information is lost information. You cannot recover what was lost during the downscaling to low resolution. Basic information theory.

The Big J said...

I would suggest you first learn how to convert low quality pics to high quality ones using your method.

I had thought of this method when I was in college but never really followed through.

Rajeev Bharshetty said...

The Big J : Thanks for the comment.
Yeah I think first experimenting on Pictures is a good idea.

Rajeev Bharshetty said...

Thanks DataWraith : Will need to do more digging on the Idea and Implementation thing.