Comparison of the visual texture calculation methods by image analysis, applied to mirror and scaled carp skin

Regions of interest (ROI) representative of the visual texture of images of mirror carp Cyprinus carpio carpio and scaled carp Cyprinus carpio were taken. Red, green, blue and grayscale (R, G, B, GS) histograms of these ROI were calculated. The following methods of visual texture calculations were performed on the ROIs: 1) image energy based on histograms, 2) image entropy based on histograms, 3) image energy based on co-occurrence matrices, 4) image entropy based on co-occurrence matrices, 5) texture based on fractal dimensions, 6) texture based on texture primitives method. Calculations were performed for color and grayscale images. The identification of the smoothest and roughest ROIs depended on the method used. The largest range between the minimum and maximum values was found in the co-occurrence matrix-based entropy calculation. A close second was the texture change index (TCI) method.

Experiment of OCITN: Considering Appropriate Goal Images and Metric for One-Class Image Transformation Network

One-class classification (OCC) is a classification problem where training data includes only one class. In such a problem, two types of classes exist, seen class and unseen class, and classifying these classes is a challenge. Besides, One-class Image Transformation Network (OCITN) is an OCC algorithm for image data. In which, image transformation network (ITN) is trained. ITN aims to transform all input image into one image, namely goal image. Moreover, the model error of ITN is computed as a distance metric between ITN output and a goal image. Besides, OCITN accuracy is related to goal image, and finding an appropriate goal image is challenging. In this paper, 234 goal images are experimented with in OCITN using the CIFAR10 dataset. Experiment results are analyzed with three image metrics: image entropy, similarity with seen images, and image derivatives.

ISAR Imaging for Maneuvering Targets with Complex Motion Based on Generalized Radon-Fourier Transform and Gradient-Based Descent under Low SNR

The existing inverse synthetic aperture radar (ISAR) imaging algorithms for ship targets with complex three-dimensional (3D) rotational motion are not applicable because of continuous change of image projection plane (IPP), especially under low signal-to-noise-ratio (SNR) condition. To overcome this obstacle, an efficient approach based on generalized Radon Fourier transform (GRFT) and gradient-based descent optimal is proposed in this paper. First, the geometry and signal model for nonstationary IPP of ship targets with complex 3-D rotational motion is established. Furthermore, the two-dimensional (2D) spatial-variant phase errors caused by complex 3-D rotational motion which can seriously blur the imaging performance are derived. Second, to improve the computational efficiency for 2-D spatial-variant phase errors compensation, the coarse motion parameters of ship targets are estimated via the GRFT method. In addition, using the gradient-based descent optimal method, the global optimum solution is iteratively estimated. Meanwhile, to solve the local extremum for cost surface obtained via conventional image entropy, the image entropy combined with subarray averaging is applied to accelerate the global optimal convergence. The main contributions of the proposed method are: (1) the geometry and signal model for ship targets with a complex 3-D rotational motion under nonstationary IPP are established; (2) the image entropy conjunct with subarray averaging operation is proposed to accelerate the global optimal convergence; (3) the proposed method can ensure the imaging accuracy even with high imaging efficiency thanks to the sole optimal solution generated by using the subarray averaging and image entropy. Several experiments using simulated and electromagnetic data are performed to validate the effectiveness of the proposed approach.