Image Retrieval Technology Research Based On Local Neighborhood Rotation Right-Angle Patterns

The accurate search of specific images from huge digital image resources has been a research focus in the field of image processing for many years.It is difficult for traditional text-based image retrieval(TBIR)to achieve retrieval goals in large databases because of its high cost,time consuming,and subjectivity.In contrast,content-based image retrieval(CBIR)is very effective in dealing with diverse and complex digital image databases.In CBIR the underlying image features are used to automatically extract image content,which avoids the drawbacks of text retrieval technology,and efficiently completes the entire image retrieval process.The underlying image features include image color features,texture features and shape features,etc.The image texture features are important and prominent visual features.The traditional image retrieval algorithm based on texture features has limitations in the choice of threshold,in which gray image selection results in the partial information loss of images and search results with high retrieval accuracy can not be provided.In order to improve the retrieval performance,it is very important to seek more effective texture features to describe image information.To solve the above problems,an image retrieval method based on local neighborhood rotation right angle pattern is proposed in this paper.Firstly,on the basis of preserving the color information of the original color image,the R,G and B single color components of the image are separated by RGB color space,and the three components are respectively discretized by two-dimensional discrete wavelet transform,and the low frequency sub-band of each component is taken.Secondly,based on the VLBP pattern,local mode is constructed for the low-frequency sub-band of each color component.On the basis of the appropriate adaptive threshold,local neighborhood rotation right-angle patterns is used to calculate the value of local neighborhood rotation right-angle patterns.Feature vector is shown by histogram and dimensionality reduction is conducted based on rotation invariant and uniform LBP pattern.Finally,the feature vectors in a single image and image database are measured by feature similarity,and image retrieval results are evaluated by the average precision(ARP)and average recall(ARR).The experimental results show that the algorithm is better than the traditional algorithm in average precision,average recall rate and the retrieval effect.