Combined low-level descriptors for improving the retrieval performance

  • Authors:
  • Mohamed Eisa

  • Affiliations:
  • Computer Science Department, Mansoura University, Mansoura, Egypt

  • Venue:
  • CSECS'08 Proceedings of the 7th conference on Circuits, systems, electronics, control and signal processing
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Most of the current visual descriptors used are calculated for full images. The local image areas of interest are easily left unnoticed as the global features do not contain enough information for local discrimination. The main contributions of this paper is enhancing the matching performance by applying different kinds of visual descriptors (color, Texture, Edge) to the sub-image areas without using any type of segmentation and compare the obtained feature descriptors separately. Three feature extraction methods, which are block-based descriptors, are presented. The first one is an advanced color feature extraction derived from the modification of Stricker's method. The second one is a texture feature extraction using the Local Binary Pattern (LBP) which is invariant, fast to calculate and its efficiency originates from the detection of different micro patterns. The third one is an edge feature extraction using the Edge Histogram Descriptor (EHD) which is time-consuming as well as computationally expensive. The experimental results demonstrate that block-based feature descriptors have good performance in terms of matching efficiency and effectiveness.