A new simple yet efficient classification algorithm named as Nested Circle Boundary (NCB) is proposed in this study. This algorithm provides features from measuring the average of the sum of boundary pixels for a number of nested circles inside the texture image. It was tested on different 91 rotated texture images for 13 texture classes using Brodatz texture database. The proposed algorithm achieves 100% accuracy when it comes to rotated texture classification. The methodology of NCB algorithm is based on two phases. Phase one mainly measures the features of the 13 texture classes and the original texture images. Then, the derived textures details are stored to be compared later with the rotated texture images. In phase two, the NCB algorithm repeats the same measures conducted in phase one but on the rotated texture images. The matching process is conducted by measuring the center values of each rotated and original image and comparing the summation of the 8 neighbors of each center with those of original images so as to find the best center value. The best center values are then assumed to be the base points for the circles, where the algorithm calculates the average of the summation of its boundary pixel values. Finally, the features of rotated textures are compared with the features of original texture images so as to find the classification solution. In conclusion, the Nested Circles Boundary Algorithm (NCB) proposed in this paper achieves the highest accuracy level using a simple technique.
- rotated texture classification ; statistical methods ; digital image ; Nested circles boundary algorithm ; texture analysis ; image processing.
Available at: http://works.bepress.com/mutaz_al-debei/14/