Huang et al. introduce a new superpixel algorithm that uses spatial and color quantization to very very efficient – called USEQ. Their approach can easily be summarized in four steps: first, a spatial quantization is computed (essentially a regular grid); second, a color quantization is computed (see paper for details), third, each pixel is – individually – assigned to a label through MAP estimation where the probability of belonging to a specific superpixel is derived from the corresponding quantizations; lastly, a post-processing step merges small superpixels.
In experiments, the authors show that the proposed method is significantly faster than many other algorithms. Still, the qualitative results, Figure 1, look convincing.
What is your opinion on the summarized work? Or do you know related work that is of interest? Let me know your thoughts in the comments below: