BennyRaphael.com
Where Benny Raphael speaks to the world
Home
My Research
Personal
Contact
Others

Links
Jokes and anecdotes
... More to come

Hidden images in photos?

After I posted this message, I realized my mistake. There was a bug in my code. But I am leaving this story since it is a useful learning experience. The bug in my code was because of this: After convoluting the image using the gradient mask, the resultant has fewer rows and columns. To make it compatible with the original image matrix, I used the numpy resize function. The resize function inserted extra numbers at the beginning of the array. So the original pixels in the first row got shifted to the right. Part of the pixels of the first row got shifted to the second row and so on. This is equivalent to shearing the image. See what I got because of this operation in the story below. I was doing some experiments with images. I extracted the gradients from images and I was surprised to see "hidden" images within photos. See this image:

It was found hidden inside this image

How did it happen? Is it due to faint reflections on the camera lens? If anyone has an answer, I am eager to find out. What I did was this: Using a python program, I extracted the gradient of colors from passport size photos, then blackened the pixels having a certain threshold. A few images with different thresholds are below:

After I wrote this, I realized there was a bug in my code! Still it is fun to see what a bug has produced. This is the line that had a bug:
np.resize(cm, (ny,nx))  # resizing the image with numpy, one pixel gets shifted all the way down

Another attempt to understand myself through the eyes of a computer

See more of my experiments with images.


.. To be continued ...
 


bennyraphael.com
Last updated Jan 28, 2006.