An Efficient Low Level Features of CBIR using Wavelet, GLDM and SMOSVM Metho
Dharmendra Pandey, Shiv Pratap Singh Kushwah · International Journal of Signal Processing Image Processing and Pattern Recognition · 2017
Content Based Image Retrieval (CBIR) is a process that allows for a person to extract an image centered on a question, from a database containing an enormous amount of pictures.A very fundamental issue in designing a CBIR system is to select the image features that best represent the image contents in a database.In this research, efficient low level features of CBIR using Bi-cubic interpolation (BCI) with color coding (CC), gray level difference method (GLDM), Hu moments and four levels of discrete wavelet transform (DWT).In this paper the techniques of CBIR are discussed, analyzed and compared.BCI is used to scale the query image and database images.CC is used for color feature extraction.Apply DWT and GLDM on each level plane of an image for texture feature.Apply Hu moments are used for shape features.The experimental database performed on Corel database which contain fruit, flowers, sports, tools, facial images.Apply Sequential Minimal Optimization Support vector matching (SMOSVM) for classifying the data.The performance analysis of precision (P), execution time (T) for retrieving images.We calculate similarity distance on Euclidean distance (ED), Manhattan Distance (MD), City Block (CBD) and Canberra distance (CD).