Data-Guided Image Retrieval System under Big Data Environment: Design and Implementation

Abstract

In Bag-of-Words based image retrieval, the SIFT visual word has a low discriminative power, so false positive matches occur prevalently. The research on data-guided image retrieval system is a hot topic. Its innovation is using PCNN and ICM in image feature extraction with translation, rotation, scale and distortion invariance and good resistance to noise, the PCNN and ICM extracted features as the image texture feature is applied to image retrieval system. The main idea is to use the PCNN and the ICM process images, get corresponds to different gray level values of binary image sequence, the sequence of the entropy of each image sequence, the one dimensional feature vector as the texture feature; Then using Euclidean distance similarity calculation. The experimental results show that the method not only has strong robustness to noise, at the same time can reduce eigenvector dimension, scale, translation and rotation invariance, and can get higher retrieval rate. In the future, we decide to do more comparison experiment to verify the effectiveness of the method.