Closed-loop adaptive image segmentation

Image segmentation is an old and difficult problem. One of the fundamental weaknesses of current computer vision systems to be used in practical applications is their inability to adapt the segmentation process as real-world changes occur in the image. We present the first closed loop image segmentation system which incorporates a genetic algorithm to adapt the segmentation process to changes in image characteristics caused by variable environmental conditions such as time of day, time of year, clouds, etc. The segmentation problem is formulated as an optimization problem and the genetic algorithm efficiently searches the hyperspace of segmentation parameter combinations to determine the parameter set which maximizes the segmentation quality criteria. The goals of our adaptive image segmentation system are to provide continuous adaptation to normal environmental variations, to exhibit learning capabilities, and to provide robust performance when interacting with a dynamic environment. We present experimental results which demonstrate learning and the ability to adapt the segmentation performance in outdoor color imagery.

Reliable transmission of high quality images through wireless communication channels represents a great challenge, especially for real-time applications. From one aspect, it is necessary to achieve efficient compression that maintains the good image quality without increasing the transmission bandwidth. Secondly, wireless channels impairments such as fading and interference increase the probability of errors when the image data is transmitted. This necessitates the use of efficient channel coding to protect against channel errors. However channel coding results in increasing the transmitted bit rate, for real time applications. Again, this is constrained by the usually limited available transmission bandwidth. In this paper, this dilemma is solved by adopting new unequal error protection techniques that are used to achieve the above requirements by using image adaptive combined compression and error protection. This is achieved through image segmentation and adaptive compression and channel coding according to the importance of the image segments. Two proposed techniques are introduced. In the first suggested technique, the image is partitioned into equal blocks and the data classification is performed in the DCT domain. The DCT coefficients of the different blocks are classified to different classes of importance and hence they are given different rates of error protection and compression ratios. The second technique is based on adaptive image segmentation using quadtree decomposition into unequal segments before applying the DCT. Finally, to improve the quality of received images, the Orientation Adaptive Sequential Interpolation error-concealment technique is applied to the reconstructed images. Computer simulation shows that the proposed techniques achieve high degree of robustness against channel impairments with good image quality with acceptable data rates.

Segmentation of intracranial aneurysm images acquired using magnetic resonance angiography (MRA) is essential for medical auxiliary treatments, which can effectively prevent subarachnoid hemorrhages. This paper proposes an image segmentation model based on a dense convolutional attention U-Net, which fuses deep and rich semantic information with shallow-detail information for adaptive and accurate segmentation of MRA-acquired aneurysm images with large size differences. The U-Net model serves as a backbone, combining dense block and convolution block attention module (CBAM). The dense block is composed of a batch normalization layer, an randomly rectified linear unit activation function, and a convolutional layer, for mitigation of vanishing gradients, for multiplexing of aneurysm features, and for improving the network training efficiency. The CBAM is composed of a channel attention module and a spatial attention module, improving the segmentation performance of feature discrimination and enhancing the acquisition of key feature information. Owing to the large variation of aneurysm sizes, multi-scale fusion is performed during up-sampling, for adaptive segmentation of MRA-acquired aneurysm images. The model was tested on the MICCAI 2020 ADAM dataset, and its generalizability was validated on the clinical aneurysm dataset (aneurysm sizes: < 3 mm, 3–7 mm, and > 7 mm) supplied by the Affiliated Hospital of Qingdao University. A good clinical application segmentation performance was demonstrated.

Digital images of the soybean and three species of weeds with different backgrounds were acquired for image processing. An adaptive neuro-fuzzy inference system (ANFIS) from MATLAB® 6.1 was used to obtain a preliminary separation of plant and background. To train the fuzzy inference system (FIS), red-green-blue (RGB) values were first converted to HSV color space (hue, saturation, value), and samples from plant and background were visually selected from the images to be used as inputs. These were then mapped to a binary output that indicated the presence or absence of plant. Subtractive clustering was used to generate the FIS with a total of 100 epochs. Three clusters, and therefore three rules, were found by the ANFIS. This resulted in a total of 9 membership functions. Source code for a genetic algorithm was written using the toolbox of MATLAB® 6.1 to adjust the membership functions to reduce misclassification and improve segmentation. The membership function parameters were used to generate chromosome genes. The initial population of 50 chromosomes was generated using 30% noise perturbation of the membership functions originally given by ANFIS. A final segmentation of background and plant was achieved after 10 generations. Between evaluations crossover and mutations were applied. The results showed that the adaptive neuro-fuzzy inference system and genetic algorithm segmentation system was capable of eliminating areas that were misclassified as plant and also generally the final segmentation.

Adaptive image segmentation for region-based object retrieval using generalized Hough transform

Reversible image watermarking schemes are used to protect ownership and copyrights of digital images. This paper proposes a novel reversible image watermarking scheme based on adaptive image interpolation, segmentation and additive prediction error expansion (PEE). Proposed interpolation comprises of weighted average of neighboring pixels by allocating higher and lower weights to less and more distant neighboring pixel values respectively. The proposed adaptive image interpolation focuses on detection of edges thus minimizing artifacts imposed by interpolation. The idea of embedding varying amount of watermark bits in different image segments has been explored. Simple linear iterative clustering (SLIC) based image segmentation is performed to separate very sharp, sharp, smooth and very smooth regions in image. Higher number of watermark bits are embedded in sharp regions by using additive prediction error expansion embedding technique. Simulations of proposed and existing techniques were performed on different images and compared using embedding capacity (EC), peak signal to noise ratio (PSNR), computational efficiency, image quality, mean square error (MSE), normalized cross correlation (NCC) and structural similarity index (SSIM). The experimental results show that proposed scheme achieves better results in terms of EC, PSNR, computational efficiency, image quality, MSE, NCC and SSIM as compared to existing techniques.

Segmentation of images into textural homogeneous regions is a fundamental problem in an image understanding system. Most region-oriented segmentation approaches suffer from the problem of different thresholds selecting for different images. In this paper an adaptive image segmentation based on vector quantization is presented. It automatically segments images without preset thresholds. The approach contains a feature extraction module and a two-layer hierarchical clustering module, a vector quantizer (VQ) implemented by a competitive learning neural network in the first layer. A near-optimal competitive learning algorithm (NOLA) is employed to train the vector quantizer. NOLA combines the advantages of both Kohonen self- organizing feature map (KSFM) and K-means clustering algorithm. After the VQ is trained, the weights of the network and the number of input vectors clustered by each neuron form a 3- D topological feature map with separable hills aggregated by similar vectors. This overcomes the inability to visualize the geometric properties of data in a high-dimensional space for most other clustering algorithms. The second clustering algorithm operates in the feature map instead of the input set itself. Since the number of units in the feature map is much less than the number of feature vectors in the feature set, it is easy to check all peaks and find the `correct' number of clusters, also a key problem in current clustering techniques. In the experiments, we compare our algorithm with K-means clustering method on a variety of images. The results show that our algorithm achieves better performance.

The paper presents an application of an adaptive object recognition technique to the detection and tracking of geographical features on aerial images. The paper advocates the necessity of the continuous image analysis for the classification of changing geographical features for aerial surveillance. The introduced technique includes: 1) extraction of geographical features by texture-based image analysis, 2) model learning and closed-loop model adaptation to the perceived changes in image characteristics, 3)recognition of interested target areas, and 4) a feedback reinforcement mechanism for model adaptation. Experimental results are presented for image sequences, along a path on an aerial image established for the aerial surveillance.

The authors reviewed their experience with pediatric pineal cysts to define the natural history and clinical relevance of this common intracranial finding. The authors identified all patients with pineal cysts who had been clinically evaluated at their institution over an 11.5-year interval and were < 25 years of age at the time of diagnosis. All inclusion criteria were met in 106 patients, and included repeated MR imaging as well as repeated clinical evaluation over at least a 6-month interval. The mean age at diagnosis was 11.7 +/- 7.2 years. Forty-two patients were male and 64 were female. On follow-up MR imaging evaluation at a mean interval of 3.0 years from the initial study, 98 pineal cysts had no increase in size and no change in imaging appearance. Six pineal cysts increased in size and 2 others had a change in imaging characteristics without associated growth. Younger age was associated with cyst change or growth on follow-up imaging (p = 0.02). The mean age of patients with cysts that changed or grew was 5.5 years, and the mean age of patients with stable pineal cysts was 12.2 years. Initial cyst size and appearance on MR imaging were not significant predictors of growth or change in imaging appearance at follow-up. Similarly, the patient's sex was not a significant predictor of growth or change in imaging characteristics. Follow-up imaging and neurosurgical evaluation may be considered optional in older children with pineal cysts.

Semantic segmentation techniques for remote sensing scene understanding have significantly advanced, enhancing the refined Earth observation. However, most methods highly depend on extensive annotated data, leading to performance deterioration in complex high-resolution remote sensing cross-domain scenes, where variations in image conditions and environments are prevalent. Domain adaptive semantic segmentation (DASS) has been proposed to mitigate the reliance on dense and costly annotations, typically using stagewise training. This article addresses three key challenges in existing DASS methods: 1) insufficient warmup training, limiting potential performance gains; 2) rigid pseudolabel threshold settings in self-training (ST) result in performance bottlenecks; 3) entropy-based prediction bias alone fails to effectively identify high-confidence noise early in ST. To address these issues, we propose a novel threshold-free pseudolabel learning framework, TPL-DA. During the warmup stage, we introduce a multiview bidirectional consistency learning mechanism within a teacher–student architecture. This mechanism employs a bias-free data augmentation strategy, fostering consistent bidirectional predictions in teacher–student networks, thereby enhancing domain generalization and feature robustness. Our multiscale context-enhanced prediction module further amplifies this. In the ST stage, we propose a dynamic threshold-free pseudolabel learning strategy that utilizes well-aligned feature prototypes in the feature space to guide pseudolabel generation in the probability space, eliminating the threshold constraints. In addition, we model uncertainty using relative entropy and incorporate it into the optimization objective to manage high-confidence noise. Extensive experiments on the LoveDA, Potsdam, and Vaihingen datasets demonstrate that TPL-DA consistently outperforms existing methods and popular benchmarks, significantly enhancing DASS performance across diverse cross-domain scenes.

License plate segmentation is a key technology in the process of license plate location and recognition. How to realize automatic segmentation of license plate image under complex illumination conditions has been a hot issue in intelligent transportation system (ITS). This paper deals with license plate image segmentation under a variety of lighting conditions. Based on the adaptive segmentation of license plate images by the Pulse Coupled Neural Network (PCNN), the relationship between the license plate image contrast and the PCNN iteration entropy is analyzed. An adaptive segmentation algorithm for license plate image using Deep Neural Network (DNN) to select the optimal result is proposed, and the selected segmentation image is filtered by the connected domain, which lays a foundation for subsequent license plate location, character segmentation and recognition. Simulation experiments show that the proposed algorithm performs better license plate segmentation and optimal selection for license plate images under various lighting conditions.

In this paper, the authors described methods of material granularity evaluation and a novel method of grain size determination with inline mill device diagnostics. The mill diagnostic can be carried out with vibration measurements, machine vision or infrared imaging. Milling process is an extremely energy- and cost-consuming process, thus diagnostics process should be performed with high efficiency. Method proposed in this paper is based on the online examination of the final product during the milling process using real-time digital images. Determination of the total number of the grain, size of each grain, also classification of the grains is main goal of proposed method. In the proposed method, the visible camera with lightning mounted at two assumed angles has been used, what increases the grain detection process. Proposed method uses an adaptive segmentation to match correctly the grains, the information about particles shape and context is used to optimize the grain classification process in the next step. Finally, during image processing, the simple rule-based method is used to obtain information about overall quality of the product and possible faults in the milling process based on the evaluated parameters.

Adaptive Mixture Estimation and Unsupervised Local Bayesian Image Segmentation

This paper aims to outline the General Adaptive Neighborhood Image Processing (GANIP) approach [1–3], which has been recently introduced. An intensity image is represented with a set of local neighborhoods defined for each point of the image to be studied. These so-called General Adaptive Neighborhoods (GANs) are simultaneously adaptive with the spatial structures, the analyzing scales and the physical settings of the image to be addressed and/or the human visual system. After a brief theoretical introductory survey, the GANIP approach will be successfully applied on real application examples in image restoration, enhancement and segmentation.

Image segmentation plays important role in computer vision and image processing, and the threshold segmentation is known by its powerful and efficient performance. Entropy-based algorithm and variance-based algorithm are two main threshold segmentation algorithms. In present paper, non- extensive Tsallis entropy is adopted to multi-level image thresholding. The non-extensive parameter is determined by the gray-level histogram in a self-adaptive way, which indicates the long-range correlations among pixels in an image. The segmentation results show that the proposed algorithm is superior to the well-known Kapur multi-level thresholding and variance- based Otsu algorithm in terms of PSNR (Peak Signal to Noise Ratio), SSIM (Structural Similarity) and uniformity measure. Furthermore, it is found that the superiority of the proposed algorithm becomes stable as the number of thresholds increases.