README

Algorithms implemented for "A Study on Connected Components Labeling algorithms using GPUs", published on:

Oliveira, V.M., Lotufo, R.A.: A study on connected components labeling algorithmsusing GPUs. In: SIBGRAPI. vol. 3, p. 4 (2010)

ccl_uf.cu -> new algorithm described in the paper 
ccl_naive_prop.cu -> naive label propagation
ccl_lequiv.cu -> label equivalence
ccl_gold.cu -> ground truth for verification
ccl_test.cu -> runs all 3 on the gpu (main function)
generate_dataset.py -> download dataset used in the paper
run.py -> runs ccl_test on the downloaded dataset
spiral.py -> utilitity to create images with spirals

I don’t have the sources for the CPU CCL implementation used in the paper because it is proprietary (it used SSE2, threads and hand-optimized assembly code).