Radon-Gabor barcodes for medical image retrieval

Abstract

In recent years, with the explosion of digital images on the Web, content-based retrieval has emerged as a significant research area. Shapes, textures, edges and segments may play a key role in describing the content of an image. Radon and Gabor transforms are both powerful techniques that have been widely studied to extract shape-texture-based information. The combined Radon-Gabor features may be more robust against scale/rotation variations, presence of noise, and illumination changes. The objective of this paper is to harness the potentials of both Gabor and Radon transforms in order to introduce expressive binary features, called barcodes, for image annotation/tagging tasks. We propose two different techniques: Gabor-of-Radon-Image Barcodes (GRIBCs), and Guided-Radon-of-Gabor Barcodes (GRGBCs). For validation, we employ the IRMA x-ray dataset with 193 classes, containing 12,677 training images and 1,733 test images. A total error score as low as 322 and 330 were achieved for GRGBCs and GRIBCs, respectively. This corresponds to ≈ 81% retrieval accuracy for the first hit.

Combining real-valued and binary Gabor-Radon features for classification and search in medical imaging archives

“…In this table A and E total refer to the accuracy and total error, respectively, obtained over all images in the testing samples. Also the vector dimension (VD) of the each feature vector has been also listed which can be computed as [5] M N Ng d1d2 where M and N are the size of the Radon transformed image, N g is the number of Gabor filters and d 1 × d 2 downsampling coefficients (algorithm 2). For example, for M = N = 32, N g = 12, and d 1 = d 2 = 4, we get VD= 768.…”

“…Highlighted results all use Gabor and/or Radon barcodes in some way. The proposed method, inspired by [5], [6] and [14], delivers the best result among this class of algorithms.…”

“…Method E total Camlica et al [11] 146.50 TAUbiomed [26] 169.50 Idiap [25] 178.93 FEITIJS [25] 242.46 Proposed Method 248.03 VPA SabancIUniv [25] 261.16 SVM+RBC [13] 294.83 MedGIFT [25] 317.53 GRIBC (5,16,23,23) [5] 330.36 RBC16 [10] 470.57…”

“…Radon barcodes based on Radon transform for encoding the local image information was introduced recently [11]. Using Gabor filters to extract binary codes from Radon transformed image was proposed by Nouresanesh et al [5], [6]. Camlica et al used an SVM classifier trained with LBP features derived from saliency image regions [12].…”

Content-based medical image retrieval system based on gradient orientation and edge information

“…Later the MinMax radon barcodes [22] is introduced to overcome the disadvantages of radon barcode that is many unique contents may lose by randon barcode and the local thresholding method doesn't cover the curvature of projection. Based on the barcodes some other methodologies [23][24][25] are also introduced.…”