Virtual resolution enhancement of natural images

We present here some illustrations of our method to magnify random textured images.
A schematic view of this approach is presented in the figure below.

Application to controlled textures

The initial image is a fractionaly pseudo-integrated CPC with controlled multiscaling exponents.
The original image is of size 32x32 and magnified by 5 successive x2 factors leading to 1024x1024 images.

Application to quiet Sun images

The original images are 2048x2048 images of the entire Sun in the extreme UV from the EIT telescope onboard SOHO.
A 32x32 part of the center of the initial image is magnified by 5 successive x2 factors leading to 1024x1024 images.

Application to non-natural images

Any kind of images ca be augmented using our method as it is consistent with the original image. We present here two illustrations of non-natural images augmented by a factor x32.
The first one is a 32x32 representation of the Matlab function 'peaks'.
The second one is just 4 squares of constant value in a 32x32 image.

Several augmentation solutions for one initial image

From an image like 'peaks', an inifinite number of different augmented images can be produced. Here, we have used three augmented images using three different distributions for the CPC multipliers.

As the details added are stochastic, using the same CPC parameters, an inifinite number of augmented images can be produced, we present here three of these solutions to the squres image.