Research And Implementation Of The Content-based Lace Image Retrieval System

Content-based image retrieval of lace is the specific research project of content-based image retrieval technology in the field of lace.The purpose of it is to search similar images in the image database of lace according to some expressive visual features.Traditional lace retrieval mainly relies on the manual retrieval and text retrieval,which leads to the unstable annotation information,low efficiency and unreliable retrieval results.Furthermore,there are many kinds of laces and lace patterns with rich and complicated content.Thus,the traditional lace retrieval no longer meets the requirements of the industry.In this paper,we do some research on the content-based image retrieval of laces and lace patterns from the algorithm and system implementation.We improve the performance of the retrieval algorithms by improving the feature matching,feature fusion and evaluation mechanism,and we develop the system based on the algorithm to achieve lace and lace pattern management and retrieval quickly and accurately.The primary task and contribution of this paper are as follows.1.We construct two image databases of 1008 laces and 606 lace patterns by using the technology of image acquisition,image preprocessing and Live Wire interactive image segmentation.2.We propose some lace retrieval methods based on improved texture feature by using hierarchical matching and feature fusion.Firstly three image-based methods,such as gray level co-occurrence matrix,gray level-gradient co-occurrence matrix,local binary pattern operator are used for texture feature matching separately.Then they are fused by means of geometry features and invariant moments for a match-by-level respectively.Finally those texture feature matching methods through hierarchical matching are fused for lace retrieval.Experimental results indicate the efficiency of the retrieval methods.3.We propose a lace retrieval algorithm containing multiple features fusion through hierarchical matching based on the fusion texture feature matching method.Firstly the texture image is processed by Canny operator and the Canny image is obtained.Then calculate its GGCMs(Gray Level-Gradient Co-occurrence Matrix).Finally the above extracted texture features are fused by means of geometry features and SURF(Speeded Up Robust Features)features for a match-by-level.4.We propose a lace pattern retrieval algorithm based on Live Wire interactive image segmentation and multiple features fusion through hierarchical matching.Firstly lace pattern contour needs to be extracted by Live Wire interactive image segmentation algorithm.Then calculate its invariant moments,shape features,and texture features to make a match-by-level respectively.5.We design and implement a lace retrieval system based on Android platform through the above algorithm research.The test results show that the recognition rate of the hierarchical matching algorithm can reach 98.37%,and the matching retrieval time is about 141.4ms.Thus,the experimental results demonstrate the feasibility and effectiveness of the proposed system to meet the requirements of real-time.