3D local circular difference patterns for biomedical image retrieval
- 14 Downloads
In this paper, three-dimensional local circular difference patterns (3D LCDP) and three-dimensional local circular difference wavelet patterns (3D LCDWP) are proposed for retrieval of biomedical images. The standard patterns are used to correlate gray value of center pixel with neighboring pixels. In the proposed approach, 3D volume is generated for calculating local circular difference patterns with the help of three planes obtained from original image. In case of color image, RGB channels are used as three planes and Gaussian filter banks of different resolution for gray level image. From this 3D volume, LCDP values are obtained by calculating relationship between center pixel and neighboring pixels in five different directions. Finally, feature vector is generated using histogram. The performance is evaluated using different medical databases: (i) open access series of imaging studies MRI database, (ii) International early lung cancer action program and vision and image analysis research groups CT scans, (iii) MESSIDOR-Retinal image database. The results are compared with existing biomedical image retrieval techniques by considering average retrieval precision and average retrieval rate as evaluation parameters.
Keywords3D LCDP 3D LCDWP Gaussian filter bank Image retrieval Local mesh patterns
- 12.Baby C, Chandy D (2013) Content based retinal image retrieval using dual tree complex wavelet transform. International conference on signal processing, image processing and pattern recognition, pp 195–199Google Scholar
- 36.Subash Kumar TG, Nagarajan V (2018) Local curve pattern for content-based image retrieval. Pattern Anal Appl 9:1–10Google Scholar
- 47.VIA/I-ELCAP CT lung image dataset. http://www.via.cornell.edu/databases-/lungdb.html