An Effective Image Retrieval Method Based on Fractal Dimension Using Kernel Density Estimation
Qin Zhang, Lin Qinghua, Kang Xin · 2017
Fractal coding has been proved useful for image compression, and it is also proved effective for image retrieval. In the paper, we present a statistical method called variable bandwidth kernel density estimation to analyze fractal coding parameters. Then retrieve images using the retrieval parameters constructed with this method. Experimental results show that the proposed method with a variable optimized bandwidth performs better than those with a fixed bandwidth and the histogram method both in retrieval rate and retrieval speed. In this paper, the Average Retrieval Rate (ARR) can reach 72.40%, which is more than that obtained by the existing methods.