Multi spacial-resolution images

I just had an idea.

A multi spacial-resolution image format where the file format would work in a similar way to the human eye.

Our eye only has a small area of hi-resolution sensing cells in the macular and the rest including the peripheral is lower resolution.

Problem it solves

This file format would reduce the overall file size of images/video where not all of the image needs to be in high resolution. This could facilitate images with higher than normal detail in parts and lower details in others.

Imagine a file format that could have regions of high and low resolution stores.

This would allow for a high level of detail in the areas where it matters (say a portrait with a low f-stop) having the details in the eye and skin whilst keeping a lower detail in the blurry background. Another advantage is that it could allow for a section of an image in 300dpi whilst another is in 600dpi and the remainder low res is in 100dpi.

Image regions could be blended in with a vectorised mask to keep the file size down as well as allow for more natural boundaries between hi and low resolution spaces.

This would also allow for video compression where a security camera could record a scene and keep a hi-res area around a suspected subject’s face or cars numberplate.

My initial thoughts would be to use something like frequency separation to split out a low and hi-frequency image set. The hi-frequency would only define smaller regions and their co-ordinates on the larger image.

But it could just as easily just keep a low res image and a bunch of hi-res regions in the frame.

Regular image compression ind video compression techniques would work on these images as it’s all just matrices. Moving images may even get a boost in compression as the total low resolution data would be less than the a regular video with all hi-res information.

Below image is a vast exaggeration of the final result.