Analyzing and Indexing Images for Art
imagedb2 is a collection of Python and OpenCV code for corpus-based analysis and indexing of images and tiles thereof for artistic purposes: remixing images, composing images from fragments, and other data-driven tasks. It is based on much earlier work (imagedb and ACTION) of mine.
See installation and dependency notes in the ImageDB2 README on GitHub, including notes on installing and running OpenCV 4.0.
The first goal is to build, test and fully explore several color/texture analysis tools:
- Full/gridded Lab/RGB/grayscale color-space histograms
- Lab/RGB/grayscale Gausiian pyramid downsampling
- Lab/RGB/grayscale DCT analysis
The second goal is to develop several approaches to indexing and/or hashing the spatial color data and forming dictionaries or hash tables for eventual retrieval. The data is spatial and exists at multiple overlapping scales.