An Improved Kernel Density Estimation with adaptive bandwidth selection for Edge detection

Edge detection has been one of the most active and challenging research aspects in the last three decades in the applications of pattern recognition and machine vision as well. Although, the recent research showed that there has been lots of work reported on the basics of Kernel density estimation (KDE), but the selection of proper bandwidth not yet reported in any literature work, because it was the most challenging issue on both under and over smoothed images in terms of non-parametric KDE. Realizing the aftermath, the current research delivers a well-defined KDE approach which allowed us an adaptive selection of bandwidth by the enactment of Shannon entropy for keeping the effect of low edge gradient in feature space. Furthermore, our adaptive bandwidth selection technique is validated through three measures like entropy, energy, and blur for more analytical analysis; along with the statistical control limit is applied to generated gradient images for appropriate edges. Finally, our proposed methodology is compared with state of art techniques in terms of the Figure of Merit (FOM) measure, and which confines a higher rate of accuracy in detecting edges. This is a more reliable and precise estimation.