HWVP: hierarchical wavelet packet descriptors and their applications in scene categorization and semantic concept retrieval

  • Authors:

  • Xueming Qian;Danping Guo;Xingsong Hou;Zhi Li;Huan Wang;Guizhong Liu;Zhe Wang

  • Affiliations:

  • School of Electronics and Information Engineering, Xi'an Jiaotong University, Xi'an, China 710049;School of Electronics and Information Engineering, Xi'an Jiaotong University, Xi'an, China 710049;School of Electronics and Information Engineering, Xi'an Jiaotong University, Xi'an, China 710049;School of Electronics and Information Engineering, Xi'an Jiaotong University, Xi'an, China 710049;School of Electronics and Information Engineering, Xi'an Jiaotong University, Xi'an, China 710049;School of Electronics and Information Engineering, Xi'an Jiaotong University, Xi'an, China 710049;School of Electronics and Information Engineering, Xi'an Jiaotong University, Xi'an, China 710049

  • Venue:
  • Multimedia Tools and Applications
  • Year:
  • 2014

Quantified Score

Hi-index 0.00

Visualization

Abstract

Wavelet packet transform is an effective texture analysis approach by sub-band filtering. Different texture patterns have distinctive responses to the sub-bands of wavelet packets. The responses are valuable for texture description. Utilizing all the responses of the sub-bands of different resolutions can improve texture pattern discrimination power. In this paper, effective texture descriptors based on hierarchical wavelet packet (HWVP) transform are proposed. The subtle sub-bands of wavelet packet transform improve the discrimination power of HWVP descriptors for the images in different categories. Scene categorization performances of the HWVP descriptors under various decomposition levels and wavelet bases are discussed. Performances of HWVP descriptors of global and local images with different partition patterns are also analyzed. The advantages of HWVP descriptors attribute to the following two aspects. Firstly sub-band filtering is helpful for improving the discrimination power of HWVP descriptors to capture the subtle differences of texture patterns. Secondly hierarchical feature representation makes the HWVP descriptors robust to resolution variations. Comparisons are made with some existing robust descriptors on scene categorization and semantic concept retrieval. Experimental results on the widely used OT, Scene-13, Sport Event, and TRECVID 2007 datasets show the effectiveness of the proposed HWVP descriptors.