Efficient content based image retrieval using transform and spatial feature level fusion

Last two decades have seen a rapid increase in the size of digital image collections. Content based image retrieval (CBIR) provides an efficient way to search and retrieve images from these large databases. In this paper, CBIR using feature level fusion of 2D complex Dual-tree Discrete Wavelet Transform and local binary pattern (LBP) is developed. Features are extracted from 2D CDT-DWT from YCbCr image as it provides better spatial and frequency resolution and can be employed to implement non-separable real-valued multidimensional discrete wavelet transforms. LBP texture descriptor detects efficiently various edges and points. The technique is evaluated using the database which includes 1000 images with 10 image classes. Average precision and recall values are computed using 30 query images from the image database. Experimental result shows the improvement in precision and recall over existing methods.