All Characteristics Preservation: Single Image Dehazing based on Hierarchical Detail Reconstruction Wavelet Decomposition Network

Image degraded by haze is a critical aspect in today's environment while getting a high-quality haze-free image remains an important task in computer vision. In recent year, many works have been done to improve the visibility of image taken under bad weather. Conventional designs use multiple image and/ or single image to deal with haze removal. In this paper, we use a simple but effective prior, a variation of distance (VoD) prior, to estimate the transmission map and remove haze from a single input image. The VoD prior is developed based on the idea that the outdoor visibility of images taken under hazy weather conditions seriously reduced when the distance increases. The thickness of the haze can be estimated effectively and a haze-free image can be recovered by adopting the VoD prior and the new haze imaging model. Our method is stable to image local regions containing objects in different depths. Our experiments showed that the proposed method achieved better results than several state-of-the-art methods, and it can be implemented very quickly. Our method due to its fast speed and the good visual effect is suitable for real-time applications. This work confirms that estimating the transmission map using the distance information instead the color information is a crucial point in image enhancement and especially single image haze removal.

DSEGAN: Detail and structure enhanced generative adversarial network for fundus image enhancement.

Haze removal is useful in computational photography and computer vision applications. Single image haze removal is still a challenge for real-time embedded systems. In general, dark channel prior approach can effectively remove haze from a single image. However, the halo effect is the major problem of dehazing. In this study, adaptive filtering, which is easy for VLSI design, is introduced to refine the transmission map and improve the result of haze removal. A hardware architecture for our haze removal method is proposed to achieve the real-time requirement. A specific hardware component is designed to compute lengths of the paths and adaptive structuring element (ASE). The hardware architecture for haze removal is implemented in 180 nm CMOS technology and occupies 5.46 mm 2 with 143 K bit on-chip memory. The design can operate at 177MHz and support for HD (1280×720) 192 frame/s. Experiment results show that the hardware implementation is suitable in embedded system for real-time HD applications.

With the rapid development of deep learning, face forgery detection methods have also achieved remarkable progress. However, most methods suffer significant performance degradation on low-quality compressed face images. It is due to: (a) The image artifacts will be blurred in the process of image compression, resulting in the model learning insufficient artifact traces; (b) Low-quality images will introduce a lot of noise information, and minimizing the training error causes the model to absorb all correlations in the training dataset recklessly, leading to the over-fitting problem. To solve the above problems, we consider learning domain invariant representations to inscribe the correct relevance, i.e., artifacts, to improve the robustness of low-quality images. Specifically, we propose a novel face forgery detector, called DIFLD. The model has the following components: (1) a high-frequency invariant feature learning module(hf-IFLM), which effectively retrieves the blurred artifacts in low-quality compressed images; and (2) a high-dimensional feature distribution learning module(hd-FDLM), that guides the network to learn more about the consistent features of distribution. With the above two modules, the whole framework can learn more discriminative correct artifact features in an end-to-end manner. Through extensive experiments, we show that our proposed method is more robust to image quality variations, especially in low-quality images. Our proposed method achieves a 3.67% improvement over the state-of-the-art methods on the challenging dataset NeuralTextures.

The Discrete Wavelet Transform (DWT) is a widely adopted technique in medical image compression for its ability to capture spatial and frequency characteristics. Clinical diagnosis may be impacted by distortions and the loss of important details caused by DWT's inability to preserve phase information, which is crucial for keeping the alignment of edges and textures in images. By incorporating a trained Autoencoder to learn and preserve essential image features for improved reconstruction, this limitation is addressed effectively. The Autoencoder comprises an encoder with convolutional layers using non-separable filters to enforce orthogonality, and a decoder with trans-posed convolutions for image reconstruction. JPEG2000 was employed as the compression technique, with the proposed method achieving a similar compression ratio to traditional DWT, indicating no compromise in efficiency. The results show that the enhanced DWT with autoencoder significantly out-performs the traditional DWT method, achieving up to 61.90% improvement in Peak Signal-to-Noise Ratio (PSNR), thereby reducing distortions and preserving critical image details more effectively. This improvement is crucial for maintaining the integrity and diagnostic quality of medical images, ensuring that essential features are accurately represented.

Gaussian noise is an important problem in computer vision. The novel methods that become popular in recent years for Gaussian noise reduction are Bayesian techniques in wavelet domain. In wavelet domain, the Bayesian techniques require a prior distribution of wavelet coefficients. In general case, the wavelet coefficients might be better modeled by non-Gaussian density such as Laplacian, two-sided gamma, and Pearson type VII densities. However, statistical analysis of textural image is Gaussian model. So, we require flexible model between non-Gaussian and Gaussian models. Indeed, Gumbel density is a suitable model. So, we present new Bayesian estimator for Gumbel random vectors in AWGN (additive white Gaussian noise). The proposed method is applied to dual-tree complex wavelet transform (DT-CWT) as well as orthogonal discrete wavelet transform (DWT). The simulation results show that our proposed methods outperform the state-of-the-art methods qualitatively and quantitatively.

Under the weather of haze, fog, or smoke, outdoor images show poor visibility and low contrast. Low contrast results in the difficulty for carrying out basic local feature (e.g., interest points and edges) detection algorithms, which are necessary procedures in some computer vision applications. Hence, increasing contrast of degraded images is very important since it is helpful in finding more distinct features from haze images. However, few single image haze removal methods can simultaneously achieve clear visibility, sufficiently high contrast, and simplicity. In this paper, we propose an intuitive and effective method, called the depth-based contrast stretching transform (DCST), to simultaneously obtain clear visibility and enhance contrast of a single haze gray image. The DCST stretches the contrast of haze images based on the coarse depth layers of scenes. Our method is simple and almost real time and can be extended to color images. We analyze in detail that the image stretched by the DCST has a higher local contrast than the image recovered via the physical-based model. Experiments demonstrate that images stretched by the DCST have excellent visibility and contrast compared with a few existing algorithms. Compelling performance is also presented by comparing the proposed method with other representative methods in the application of local feature detection.

Low-dose computed tomography (LDCT) has promising potential for dose reduction in medical applications, while suffering from low image quality caused by noise. Therefore, it is in urgent need for developing new algorithms to obtain high-quality images for LDCT. This study tries to exploit the sparse and low-rank properties of images and proposes a new algorithm based on subspace identification. The collection of transmission data is sparsely represented by singular value decomposition and the eigen-images are then denoised by block-matching frames. Then, the projection is regularized by the correlation information under the frame of prior image compressed sensing (PICCS). With the application of a typical analytical algorithm on the processed projection, the target images are obtained. Both numerical simulations and real data verifications are carried out to test the proposed algorithm. The numerical simulations data is obtained based on real clinical scanning three-dimensional data and the real data is obtained by scanning experimental head phantom. In simulation experiment, using new algorithm boots the means of PSNR and SSIM by 1 dB and 0.05, respectively, compared with BM3D under the Gaussian noise with variance 0.04. Meanwhile, on the real data, the proposed algorithm exhibits superiority over compared algorithms in terms of noise suppression, detail preservation and computational overhead. The means of PSNR and SSIM are improved by 1.84 dB and 0.1, respectively, compared with BM3D under the Gaussian noise with variance 0.04. This study demonstrates the feasibility and advantages of a new algorithm based on subspace identification for LDCT. It exploits the similarity among three-dimensional data to improve the image quality in a concise way and shows a promising potential on future clinical diagnosis.

BackgroundTo study the impact of image quality on quantitative measurements and the frequency of segmentation error with optical coherence tomography angiography (OCTA).MethodsSeventeen eyes of 10 healthy individuals were included in this study. OCTA was performed using a swept-source device (Triton, Topcon). Each subject underwent three scanning sessions 1–2 min apart; the first two scans were obtained under standard conditions and for the third session, the image quality index was reduced using application of a topical ointment. En face OCTA images of the retinal vasculature were generated using the default segmentation for the superficial and deep retinal layer (SRL, DRL). Intraclass correlation coefficient (ICC) was used as a measure for repeatability. The frequency of segmentation error, motion artifact, banding artifact and projection artifact was also compared among the three sessions.ResultsThe frequency of segmentation error, and motion artifact was statistically similar between high and low image quality sessions (P = 0.707, and P = 1 respectively). However, the frequency of projection and banding artifact was higher with a lower image quality. The vessel density in the SRL was highly repeatable in the high image quality sessions (ICC = 0.8), however, the repeatability was low, comparing the high and low image quality measurements (ICC = 0.3). In the DRL, the repeatability of the vessel density measurements was fair in the high quality sessions (ICC = 0.6 and ICC = 0.5, with and without automatic artifact removal, respectively) and poor comparing high and low image quality sessions (ICC = 0.3 and ICC = 0.06, with and without automatic artifact removal, respectively).ConclusionsThe frequency of artifacts is higher and the repeatability of the measurements is lower with lower image quality. The impact of image quality index should be always considered in OCTA based quantitative measurements.

In the enhancement of low-light images, noise, low contrast, color distortion, and loss of details have remained great issues that have not been overcome by computer vision. The traditional products like gamma correction, CLAHE, and histogram equalization are used to improve the brightness, but they are incapable of finding the balance between boosting and reducing noises. Recent Retinex, GAN and encoder decoder processes show improvements but are incapable of preserving fine details of natural color restoration. The current paper suggests a hybrid model that comprises SwinIR-denoising, UNet reconstruction using attention-focused skip connections, and guided filtering with Laplacian decomposition. It is stated that the model trained in the LOL dataset achieved PSNR of 20.21 dB, SSIM of 0.774 and MSE of 0.0134 in a 58 -minute training period and thus it has good brightness correction abilities, detail preservation and realistic texture reproduction in low-light images.

This paper proposes a new methodology to address the image quality problem encountered as the use of an unmanned aerial vehicle (UAV) in the field of bridge inspection increased. When inspecting a bridge, the image obtained from the UAV was degraded by various interference factors such as vibration, wind, and motion of UAV. Image quality degradation such as blur, noise, and low-resolution is a major obstacle in utilizing bridge inspection technology based on UAV. In particular, in the field of bridge inspection where damages must be accurately and quickly detected based on data obtained from UAV, these quality issues weaken the advantage of using UAVs by requiring re-take of images through re-flighting. Therefore, in this study, image quality assessment (IQA) based on local blur map (LBM) and image quality enhancement (IQE) using the variational Dirichlet (VD) kernel estimation were proposed as a solution to address the quality issues. First, image data was collected by setting different camera parameters for each bridge member. Second, a blur map was generated through discrete wavelet transform (DWT) and a new quality metric to measure the degree of blurriness was proposed. Third, for low-quality images with a large degree of blurriness, the blind kernel estimation and blind image deconvolution were performed to enhance the quality of images. In the validation tests, the proposed quality metric was applied to material image sets of bridge pier and deck taken from UAV, and its results were compared with those of other quality metrics based on singular value decomposition (SVD), sum of gray-intensity variance (SGV) and high-frequency multiscale fusion and sort transform (HiFST) methods. It was validated that the proposed IQA metric showed better classification performance on UAV images for bridge inspection through comparison with the classification results by human perception. In addition, by performing IQE, on average, 26% of blur was reduced, and the images with enhanced quality showed better damage detection performance through the deep learning model (i.e., mask and region-based convolutional neural network).

ClarityDiffuseNet: Enhancing fundus image quality under black shadows with diffusion model-based research

The objective of getting a good quality of image plays an important role in applications related to image processing. To achieve this, the quality of image should be monitored dynamically and by applying optimized algorithms. Thus, the zoomed image got using algorithms is sharper compared to other methods. Hence keeping all this in mind on this source of information, discrete wavelet transform (DWT) with various interpolation techniques had been applied upon variances to obtain their values. The proposed method gives better PSNR compared to other methods. Thus in this paper, an algorithmic approach is used to zoom a given image in wavelet domain and to get a sharper image using various interpolation techniques.

In this paper, we propose a simple but effective image prior - dark channel prior to remove haze from a single input image. The dark channel prior is a kind of statistics of the haze-free outdoor images. It is based on a key observation - most local patches in haze-free outdoor images contain some pixels which have very low intensities in at least one color channel. Using this prior with the haze imaging model, we can directly estimate the thickness of the haze and recover a high quality haze-free image. Results on a variety of outdoor haze images demonstrate the power of the proposed prior. Moreover, a high quality depth map can also be obtained as a by-product of haze removal.

As a local filter, the guided image filtering (GIF) suffers from halo artifacts. To address this issue, a novel weighted aggregating strategy is proposed in this paper. By introducing the weighted aggregation to GIF, the proposed method called WAGIF can achieve results with sharp edges and avoid/reduce halo artifacts. More specifically, compared to the weighted guided image filtering and the gradient domain guided image filtering, the proposed method can achieve both fine and coarse smoothing results in the flat areas while preserving edges. In addition, the complexity of the proposed approach is O(N) for an image with N pixels. It is demonstrated that the GIF with weighted aggregation performs well in the fields of computational photography and image processing, including single image detail enhancement, tone mapping of high-dynamic-range images, single image haze removal, etc.