Combining PCNN with color distribution entropy and vector gradient in feature extraction
Cheng Yang, Xiaodong Gu · 2012
In this paper, the simplified Pulse-Coupled Neural Network (PCNN) model, widely used in image processing, is used to extract image features for image retrieval. These features include PCNN-segmentation-based color information and PCNN-gradient-based texture. On one hand, considering the spatial distribution of colors, we combine the color distribution entropy with the simplified PCNN. On the other hand, we also make use of the texture features of images produced by gradient images. Experimental results show that our method performs better than Improved Color Distribution Entropy (ICDE), Block Difference of Inverse Probabilities (BDIP), PCNN-Global Icon (PCNN-GI) and Normalized Moment of Inertia (Nmi) method respectively for recall-precision and ANMRR index.