Real Complex Images Research Articles

Ensemble CART surrogate-assisted automatic multi-objective rough fuzzy clustering algorithm for unsupervised image segmentation

Multi-objective clustering algorithms (MOCAs) are popular in unsupervised image segmentation due to their merit of meeting multiple segmentation requirements and the prospect of automatically estimating the number of clusters. However, most of them suffer from high time costs and are easily to be influenced by the uncertainty when handling real complex images. To address these issues, we propose an ensemble classification and regression tree (CART) surrogate-assisted automatic multi-objective rough fuzzy clustering (ECS-AMRFC) algorithm for unsupervised image segmentation. Firstly, a cluster medoid-based encoding scheme is employed to represent solutions with different number of clusters and meanwhile lessen the length of encoding. Then, we design an ensemble CART as the surrogate model to significantly reduce the computational burden. Moreover, a surrogate model management strategy is proposed to accelerate the optimization and enhance the quality of surrogate modeling. To handle the uncertainty in data, we extend the rough fuzzy clustering into MOCAs and construct three complementary objective functions to seek proper cluster medoids from multiple perspectives. In addition, the Gaussian kernel is introduced into the objective functions to handle image pixels that cannot separate linearly in the feature space. Finally, a kernelized rough fuzzy clustering validity index is defined to automatically select the optimal solution with no requirements of any prior knowledge. Experiments show that ECS-AMRFC not only identifies appropriate number of clusters on different kinds of images, but also obtains better segmentation results than state-of-the-art rough fuzzy clustering algorithms and automatic MOCAs.

A Hybrid SAR/ISAR Approach for Refocusing Maritime Moving Targets with the GF-3 SAR Satellite

Due to self-motion and sea waves, moving ships are typically defocused in synthetic aperture radar (SAR) images. To focus non-cooperative targets, the inverse SAR (ISAR) technique is commonly used with motion compensation. The hybrid SAR/ISAR approach allows a long coherent processing interval (CPI), in which SAR targets are processed with ISAR processing, and exploits the advantages of both SAR and ISAR to generate well-focused images of moving targets. In this paper, based on hybrid SAR/ISAR processing, we propose an improved rank-one phase estimation method (IROPE). By using an iterative two-step convergence approach in the IROPE, the proposed method achieves accurate phase error, maintains robustness to noise and performs well in estimating various phase errors. The performance of the proposed method is analyzed by comparing it with other focusing algorithms in terms of processing simulated data and real complex image data acquired by Gaofen-3 (GF-3) in spotlight mode. The results demonstrate the effectiveness of the proposed method.

NO-REFERENCE QUANTITATIVE ASSESSMENT OF GENERALIZED CONTRAST FOR COMPLEX IMAGES

Context. Nowadays the task of automatically measuring of image quality in real time is extremely relevant for the vast majority of practical applications. No-reference quantitative assessment of image quality is one of the most pressing and difficult problems of image processing. Generalized contrast is the most important quantitative characteristic which determines the objective quality of the image. Currently, the development of new effective methods of no-reference measuring of generalized contrast for complex image in automatic mode which have the level of computing costs, which are acceptable to implement the processing in real time, is one of the most urgent tasks of image pre-processing.Objective. Development of new histogram-based method for no-reference measurement of generalized contrast of complex images on the basis of the mean value for all contrast values of all pairs of image elements (objects and background) for various definitions of contrast kernel.Method. Analysis of known approaches to measurement of a local contrast of the image elements, of known methods of the quantitative assessment of generalized contrast of complex images as well of the results of experimental research for a series of complex real and test images allowed to reveal inherent patterns (accordance to basic requirements to the definition of contrast, the nature and the dynamic of contrast changes at the linear transformations of the brightness scale), which are manifested depending on the use of the different definitions of the contrast kernels and the metrics of generalized contrast of images.Results. New histogram-based method for no-reference measurement of generalized contrast for complex images is proposed. No-reference contrast metrics for the histogram-based measuring of generalized contrast of complex images on the basis of the average contrast of image elements for different definitions of contrast kernel is proposed.Conclusions. Proposed no-reference metrics on the basis of the average contrast of image elements for proposed contrast kernels allow providing accurate quantitative assessment of generalized contrast of the real complex images and enable to evaluate (predict) with reasonable accuracy the perceived image quality at carrying out of subjective (qualitative) expert estimates.

HISTOGRAM-BASED METHOD FOR NO-REFERENCE MEASUREMENT OF GENERALIZED CONTRAST OF COMPLEX IMAGES

Background. Nowadays the task of automatically measuring of image quality in real time is extremely relevant for the vast majority of practical applications. No-reference quantitative assessment of image quality is one of the most pressing and difficult problems of image processing. Generalized contrast is the most important quantitative characteristic which determines the objective quality of the image. Currently, the development of new effective methods of no-reference measuring of generalized contrast for complex image in automatic mode with the level of computing costs, which are acceptable to implement the processing in real time, is one of the most urgent tasks of image preprocessing. Objective. Development of new histogram-based method for no-reference measurement of generalized contrast of complex (multi-element) images based on the average contrast of image elements (objects and background) for different definitions of contrast kernel. Methods. Analysis of known approaches to measurement of a local contrast of the image elements, of known methods of the quantitative assessment of generalized contrast of complex images as well as of the experimental research results for a series of complex real and test images allowed revealing inherent patterns (accordance to basic requirements to the definition of contrast, the nature and the dynamic of contrast changes at the linear transformations of the brightness scale), which are manifested depending on the use of the different definitions of the contrast kernels and the metrics of generalized contrast of images. Results. No-reference contrast metrics for the histogram-based measuring of generalized contrast of complex images based on the average contrast of image elements for different definitions of contrast kernel is proposed. Conclusions. Proposed no-reference metrics based on the average contrast of image elements for proposed contrast kernels allow providing accurate quantitative assessment (measurement) of generalized contrast of the real complex images and enable to evaluate (predict) with reasonable accuracy the perceived image quality at carrying out of subjective (qualitative) expert estimates.

Image Mosaicing based on Neural Networks

The main concept behind image mosaic is image registration. In image mosaicing several overlapping images are assembled in order to constitute one panoramic image. In this paper a new feature-based approach will be presented for automated image to image registration and mosaicing. The proposed method is implemented on real complex images. The proposed method is based on five main steps. First, the Harris algorithm is used to extract the feature points in the reference and sensed images. Second, feature matching is established using the Euclidean distance of the signature vectors obtained using pulse coupled neural network (PCNN). Third, transformation parameters are obtained using the least-square rule based on general affine transformation. Fourth, the image resampling and transformation are performed using bilinear interpolation to get the registered image. Finally, the mosaicing image is obtained. Experimental results show that the proposed algorithm shows excellent results when applied and tested on real complex images. General Terms Image Processing, image Registration, image mosaicing and image stitching

Lane Markers Detection Based on Consecutive Threshold Segmentation

This paper proposed a simple and robust lane markers detection method for intelligent vehicle navigation. It needs not calculate inverse perspective map. The method uses multiple threshold segmentation instead of single threshold segmentation. And straight and curve lane markers are directly extracted in Run-Length accumulation (RLA) images. It performs well in various complex conditions and costs less than 50 ms for a 352 by 288 image. Experiments on many kinds of real complex image sequences demonstrate the effectiveness and efficiency of the proposed method.