The box dimension for researching similarity in fractal image coding

Fractal image compression is a lossy image coding method using partitioned iterated function systems (PIFS). Compared with rapid decompression algorithms, the compression process is extremely time-consuming, so how to speed up the compression procedure remains a challenging issue. The most common solution involves classification of domain and range blocks according to features, after which matches across class boundaries are excluded. We compare two feature vector methods - mass center and box dimension. Experimental results demonstrate the improvements in compression performance.