Channel adaptive underwater image enhancement algorithm based on pixel level

Adaptive Fractional Order Differential Image Enhancement Algorithm Based on Image Complexity

Underwater image enhancement is one of the most essential and fundamental tasks in ocean investigations recent years. In this paper, we made an attempt to develop one adaptive underwater image enhancement approach with the help of the virtual retina model and the image quality assessment (IQA). The virtual retina model, which yields comparatively high correlation with the human vision system, is first taken to achieve simultaneous ambiguity removing and detail enhancing of single image due to the specific mechanisms of different retinal sub-layers. After this, an adaptive image enhancement strategy is taken with one kind of no-reference image quality assessment based Patch Discrete Cosine Transform (PDCT), which indicates whether the image patches are naturally uniform or not. It is shown in the simulation experiment that the proposed approach could achieve great performances in both robustness and effectiveness, with good behaviors in the vision effects and precision for the underwater images.

Environment awareness technology is one of the most critical technologies for autonomous vehicles, and traffic sign recognition is an important branch of environment awareness system, which has important research significance for ensuring traffic safety. To solve the problems of easy omission and inaccurate positioning for traffic sign detection under complex illumination conditions, there are two main innovations in this paper: firstly, an adaptive image enhancement algorithm was proposed to improve the image quality under complex illumination conditions; Secondly, a novel and lightweight attention block named Feature difference (FD) model was proposed to detect and recognize traffic signs. Unlike state-of-the-art attention models, which utilizes the difference between two feature maps to generate the attention mask. In this work, the single-stage target detection algorithm SSD was selected as the basic network, the backbone network was set as ResNet and VGG respectively and FD module was added. A large number of experiments were carried out to evaluate the optimization effect of FD module. The following conclusions can be drawn: image enhancement algorithms can provide better image samples; The FD model can be integrated into most of CNNs just by adding some shortcut connections, which does not introduce any additional parameters and layers; The average detection accuracy with FD module is 1.80% higher and the average recall rate was 1.51% higher than that without FD module, which has no great influence on the running speed. The FD module has better enhancement effect for dark image detection; The detection accuracy of the FD module is slightly better than other attention modules including BAM, CBAM and SE; The FD model can be used to evaluate the effect of the convolutional operation on features, which could help to find convolutional layers that have little effect on the accuracy of recognition for the network pruning.

How to capture and enhance the target from an image with complex texture and edge features is studied in this paper. For the traditional image enhancement methods can not split different kinds of the textures and edges of an image efficiently, a novel adaptive enhancement algorithm based on nonsubsampled contourlet transform (NSCT) is projected. Firstly, by using the direct spatial multiplication of NSCT coefficients of each same directional subband at adjacent highpass scales, the inter-scale spaces are constructed and the optimal geometrical directions of each matrix region are calculated in the proposed spaces. Experimental results show that, in NSCT domain, the geometrical direction information which locates in the inter-scale spaces can represent the different detail features exactly of an image. Therefore, the optimal geometrical direction information as a rule to determine what kind of edge and texture features would be captured and enhanced is employed. After comparing with existing enhancement method, the proposed approach achieves better results, not only clearly identifies target edges and textures from complex background, but also enhances them with better visualization.

Vision-based underwater exploration is crucial for marine research. However, the degradation of underwater images due to light attenuation and scattering poses a significant challenge. This results in the poor visual quality of underwater images and impedes the development of vision-based underwater exploration systems. Recent popular learning-based Underwater Image Enhancement (UIE) methods address this challenge by training enhancement networks with annotated image pairs, where the label image is manually selected from the reference images of existing UIE methods since the groundtruth of underwater images do not exist. Nevertheless, these methods encounter uncertainty issues stemming from ambiguous multiple-candidate references. Moreover, they often suffer from local perception and color perception limitations, which hinder the effective mitigation of wide-range underwater degradation. This paper proposes a novel NUAM-Net (Novel Underwater Image Enhancement Attention Mechanism Network) that addresses these limitations. NUAM-Net leverages a probabilistic training framework, measuring enhancement uncertainty to learn the UIE mapping from a set of ambiguous reference images. By extracting features from both the RGB and LAB color spaces, our method fully exploits the fine-grained color degradation clues of underwater images. Additionally, we enhance underwater feature extraction by incorporating a novel Adaptive Underwater Image Enhancement Module (AUEM) that incorporates both local and long-range receptive fields. Experimental results on the well-known UIEBD benchmark demonstrate that our method significantly outperforms popular UIE methods in terms of PSNR while maintaining a favorable Mean Opinion Score. The ablation study also validates the effectiveness of our proposed method.

The marine aquaculture fishing industry has caused many problems such as low work efficiency and hidden safety hazards due to low automation. Research on marine fishing robots that can replace humans for autonomous operations has excellent significance and prospects. Aiming at the problems of uneven lighting, poor visibility, and many magazines in the underwater environment, the underwater fishing robot can quickly and autonomously detect marine products and other targets. This paper implements an underwater image enhancement algorithm that can be deployed on a small AI system and an object detection algorithm based on YOLOv5 (You Only Look Once Version 5). The algorithm includes using the limited contrast adaptive histogram equalization image enhancement technology to solve the image quality problems of blue-green and blurry underwater images and then using the YOLOv5 object detection network to detect and locate underwater creatures. Experimental results show that the algorithm can effectively solve poor underwater image quality and unclear targets and rapidly detect seafood targets. The detection accuracy of this algorithm can reach 85%. It has been applied to the underwater fishing robot independently developed by the team and deployed on the Jetson Xavier NX small AI system. The detection accuracy can reach 80%, and the detection speed can reach 30FPS.

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.

This work proposes an adaptive image enhancement (AIE) method, is used to enhance handwritten document images. An AIE method is proposed to denoise handwritten documents in a wavelet domain, which differs from other methods in two aspects: Firstly, modified contrast limited adaptive histogram equalization (CLAHE) is used to equalize the contrast of an image by cutting the histogram at some threshold, and then equalization is used. Secondly, the image is improved by using directional transform enhancing for foreground and interfering strokes, respectively. In the experimental result, the proposed method not only removes the interfering strokes or visible watermarks in the background information, but also significantly increases the readability of handwritten document images.

An adaptive 3D imaging technique with phase compensation for atmospheric radar imaging is presented. For imaging the atmospheric structure in detail, the advection by the background wind causes phase rotation that prevents coherent integration for improving signal to noise ratio. Synthetic aperture radar (SAR) technique is one general approach to a target drifting with a known motion vector but the method requires the target to reflect waves coherently. We propose a compensation method for phase rotation caused by the advection to apply adaptive high-resolution imaging algorithms to this problem. By making correlation matrix after the phase compensation determined by measured wind velocity to received signal, an adaptive imaging such as Capon's method can be performed. This phase compensated signal, however, should be made for each points in the imaging space which takes large computation cost. To reduce the computation cost, we also introduce an approximation technique that makes the compensated signal represent certain region in the space of interest whose dimension is determined by the limit of approximation error of signal phases. (4 pages)

Turbidity-graded adaptive underwater image enhancement

In underwater imagery, issues such as non-uniform illumination, blurriness, and low contrast are prevalent, significantly impacting the quality of captured images. In recent years, numerous researchers have delved into underwater image processing. Due to the intricacies of underwater environments, low-light images have different requirements compared to well-illuminated ones. However, existing algorithms often struggle to address the non-uniform illumination issues stemming from various lighting conditions in underwater settings. They also lack the capability to adaptively enhance underwater images with varying brightness. To tackle these challenges, we propose an adaptive illumination enhancement method for underwater images. This algorithm offers the capability to adaptively enhance underwater images suffering from detail blurriness based on their original brightness. Furthermore, it dynamically adjusts the parameters of the gamma function using the image's illumination component to augment color contrast. Experimental results demonstrate that our approach outperforms other algorithms, as evidenced by superior scores in UIQM metric. It effectively addresses edge blurriness and non-uniform illumination issues prevalent in underwater images captured under varying lighting conditions.

Coal gangue separation serves as an effective method to enhance coal quality, and machine vision-based coal gangue detection methods hold significant research and application potential. Industrial dust leads to unclear imaging and affects detection accuracy. However, there is currently little research on the impact of dust. To achieve high visual detection performance of coal and gangue under dust conditions, this study researched visual characteristics and detection methods under smoke-simulated dust conditions. Firstly, we analyze the influence of smoke concentration on the coal gangue images captured by a line array camera, showing that the higher the smoke concentration, the lower the image brightness and clarity. Subsequently, we proposed an adaptive image enhancement algorithm MSRCP-Ada-BR by integrating an adaptive brightness regularization method into the MSRCP algorithm. Experimental results demonstrate that this algorithm significantly enhances the brightness and clarity of coal gangue images across varying smoke concentrations, aligning the pixel distribution of the enhanced images more closely with that of smoke-free images. Ultimately, by integrating the MSRCP-Ada-BR into the input of YOLOx network, we construct a coal gangue detection model suitable for dusty environments. Compared to the original YOLOx model, the detection model integrated with MSRCP-Ada-BR achieves a 5.59% improvement in accuracy.

The paper presents a new Adaptive Gain calculating approach when using the Adaptive Image Enhancement algorithm based on Wavelet Transform. The basic technique is to select two different thresholds which divide the input into three parts after the wavelet coefficients are normalized. For the wavelet coefficients less than the smaller threshold, just make the output zero. And this provides a de-noise effect. The output remains unchanged when the wavelet coefficients are greater than the larger one. For the last part, find a function to make the output figure S shape. So the algorithm can give a clear contrast and the function is the key. The final goal is to use this method in the process of Online Vision Measure, we have chosen the TI TMS320 DM642 Digital Signal Processor or DSP because of its powerful multimedia processing capability. What's more, the TI corporation has provided a variety of software develop libraries as well as the 3rd party tools. All such tools make the development more rapid and convenient. After the technique this paper provides is implemented on DSP, a series of optimization will be performed to make it suitable for industry real-time usage.

The traffic block port monitors and manages the road traffic by shooting and recording the motor vehicles. However, due to the complex factors such as shooting angle, light condition, environmental background, etc., the recognition rate of license plate is not high enough. High light and low light under complex lighting conditions are symmetry problems. This paper analyzes and solves the low light problem in detail, an image adaptive enhancement algorithm under low light conditions is proposed in the paper. The algorithm mainly includes four modules, among which, the fast image classification module uses the deep and separable convolutional neural network to classify low-light images into low-light images by day and low-light images by night, greatly reducing the computation burden on the basis of ensuring the classification accuracy. The image enhancement module inputs the classified images into two different image enhancement algorithms and adopts the idea of dividing and ruling; the image quality evaluation module adopts a weighted comprehensive evaluation index. The final experiment shows that the comprehensive evaluation indexes are all greater than 0.83, which can improve the subsequent recognition of vehicle face and license plate.

Current camera image and signal processing pipelines (ISPs), including deep-trained versions, tend to apply a single filter that is uniformly applied to the entire image. This is despite the fact that most acquired camera images have spatially heterogeneous artifacts. This spatial heterogeneity manifests itself across the image space as varied Moire ringing, motion blur, color bleaching, or lens-based projection distortions. Moreover, combinations of these image artifacts can be present in small or large-pixel neighborhoods within an acquired image. Here, we present a deep reinforcement learning model that works in learned latent subspaces and recursively improves camera image quality through patch-based spatially adaptive artifact filtering and image enhancement. Our Recursive Self-Enhancement Reinforcement Learning(RSE-RL) model views the identification and correction of artifacts as a recursive self-learning and self-improvement exercise and consists of two major sub-modules: (i) The latent feature subspace clustering/grouping obtained through variational auto-encoders enabling rapid identification of the correspondence and discrepancy between noisy and clean image patches. (ii) The adaptive learned transformation is controlled by a soft actor-critic agent that progressively filters and enhances the noisy patches using its closest feature distance neighbors of clean patches. Artificial artifacts that may be introduced in a patch-based ISP are also removed through a reward-based deblocking recovery and image enhancement. We demonstrate the self-improvement feature of our model by recursively training and testing on images, wherein the enhanced images resulting from each epoch provide a natural data augmentation and robustness to the RSE-RL training-filtering pipeline. Our code is publicly available at https://github.com/CVC-Lab/rse_rl.git .