Image Pixel Visualization in Blender

Generates 3D depictions of images in Blender. Intended for use in computer-vision education and such.

Samples

My living room, one channel

My living room after being passed through a Gaussian filter

A visualization of a 21x21 Gaussian kernel

A visualization of a 21x21 Gabor kernel

Dependencies

This depends on the Python Image Library for image parsing.

pip install pillow

Usage

sh draw_image.sh my_image.png

Without the command line, you can copy and paste the contents of json_image_to_blender.py into Blender, update the path to pixels.json, and run.

This is intended to work only with grayscale images. But, by messing with the image_to_json.py file, you can do multi-channel visualizations easily enough.

The blender_filter_sampler.py file generates json for common convolution kernels. Open the file, comment/uncomment whatever functions you want, and then run python blender_filter_sampler.py > pixels.json.

Warning: The script cannot handle images larger than ~50,000 pixels. Scale input images down before running.