Improved Similar Images Retrieval: Dynamic Multi-feature of Fusion a Method with Texture Features

Rapid advances in digital image technology contributes huge amount of images in the many fields like digital libraries, medical imaging, Web image searching, etc. This chapter attempts to present novel approach in CBIR, which is based on the combined texture features. This approach mainly aims at designing the effectiveness of the image retrieval system. Thus, in this chapter, multiple features method has been proposed in order to create a novel hybrid features by 4Dir-GLCMEDGE method. This method majorly has three process flows, (a) a process to find the four diagonal texture features (b) a process to find the local binary pattern (LBP) features (c) It is a fusion of diagonal texture features and edge features. This fusion aims at combining and reducing multiple image features into single image feature. Selecting a query image as input, more similar images can be retrieved with the help of the proposed hybrid features. Using the hybrid features, based on two distance metrics such as euclidean and manhattan metics, results in more similar imges. This proposed method has been tested with Corel and Wang datasets. The test results, compared to the previous methods, are encouraging and found to be better than existing methods. The image retrieval result improvements in the new approach method over existing methods have been proved by in terms of precision and recall.