Adaptive Sharpening Research Articles

Taming Overconfident Prediction on Unlabeled Data From Hindsight.

Minimizing prediction uncertainty on unlabeled data is a key factor to achieve good performance in semi-supervised learning (SSL). The prediction uncertainty is typically expressed as the entropy computed by the transformed probabilities in output space. Most existing works distill low-entropy prediction by either accepting the determining class (with the largest probability) as the true label or suppressing subtle predictions (with the smaller probabilities). Unarguably, these distillation strategies are usually heuristic and less informative for model training. From this discernment, this article proposes a dual mechanism, named adaptive sharpening (ADS), which first applies a soft-threshold to adaptively mask out determinate and negligible predictions, and then seamlessly sharpens the informed predictions, distilling certain predictions with the informed ones only. More importantly, we theoretically analyze the traits of ADS by comparing it with various distillation strategies. Numerous experiments verify that ADS significantly improves state-of-the-art SSL methods by making it a plug-in. Our proposed ADS forges a cornerstone for future distillation-based SSL research.

An improved method for upscaling 2D images using adaptive edge sharpening and an optimized directional anisotropic diffusion filter

Image upscaling employs various polynomial techniques, including bicubic and Lanczos, to minimize computing load for various real-time applications. Polynomial interpolation, on the other hand, results in blurring artifacts in high resolution (HR) images due to edge degradation. Many other interpolation approaches, including edge-based and dictionary learning-based schemes, result in blurring in the image’s high frequency (HF) region. As a result, a novel post-processing technique is proposed that uses the Lanczos method to upscale the low resolution (LR) image. To eliminate blur along the image’s edge, the edge component of the interpolated image is retrieved. Adaptive sharpening is used to make the edges more prominent, as predicting pixels in the high variance region is challenging. After the edge is extracted, the smooth region is filtered with an optimized directional anisotropic diffusion (ODAD) filter to preserve texture details and geometric regularity. The sharpened edge image is aggregated with the ODAD filtered image to generate an improved HR (IHR) image. The IHR is then scaled up and down to reduce blurring even further, resulting in a blurred HR image (BHR). The residual image is obtained by subtracting the BHR from the IHR images, which represent the undetected HF. Finally, the residual image and the IHR image are combined to generate the final restored image, which includes all prediction loss details. The results of the experiments show that the proposed algorithm outperforms existing methodologies, particularly in terms of visuality and parameter-wise measurement.

Multimodal Image Fusion for X-ray Grating Interferometry.

X-ray grating interferometry (XGI) can provide multiple image modalities. It does so by utilizing three different contrast mechanisms-attenuation, refraction (differential phase-shift), and scattering (dark-field)-in a single dataset. Combining all three imaging modalities could create new opportunities for the characterization of material structure features that conventional attenuation-based methods are unable probe. In this study, we proposed an image fusion scheme based on the non-subsampled contourlet transform and spiking cortical model (NSCT-SCM) to combine the tri-contrast images retrieved from XGI. It incorporated three main steps: (i) image denoising based on Wiener filtering, (ii) the NSCT-SCM tri-contrast fusion algorithm, and (iii) image enhancement using contrast-limited adaptive histogram equalization, adaptive sharpening, and gamma correction. The tri-contrast images of the frog toes were used to validate the proposed approach. Moreover, the proposed method was compared with three other image fusion methods by several figures of merit. The experimental evaluation results highlighted the efficiency and robustness of the proposed scheme, with less noise, higher contrast, more information, and better details.

Degraded document image preprocessing using local adaptive sharpening and illumination compensation

Degraded document image preprocessing using local adaptive sharpening and illumination compensation

Noise-Robust MRI Upsampling Using Adaptive Local Steering Kernel

Upsampling and denoising of magnetic resonance images are conventionally performed separately, which would introduce unwanted artifacts such as blurring. To address this problem, we propose an innovative adaptive interpolation framework to achieve simultaneous image upsampling, denoising, and detail enhancement. First, local steering kernel (LSK) function is leveraged to adapt the interpolation weights according to geometric structures in the magnetic resonance (MR) images. An adaptive sharpening of the LSK weight matrix and a Rician bias correction are then adopted to remove Rician noise and enhance fine details. In this regard, the adaptive LSK extends the zero-order point estimation framework to higher orders of regression, and therefore facilitates edge preservation and detail reconstruction. The post-processing Rician correction avoids the bias caused by the asymmetry of Rician noise distributions. Experimental results using both real and synthetic clinical MR cranial images (with and without noise) demonstrated that our algorithm produced better reconstruction results than several traditional interpolation-based upsampling methods, including nearest neighbor (NN), non-local means (NLM), self-learning super resolution (SLSR), Gaussian process regression (GPR), and even comparable to four deep-learning-based methods but with less data requirements and computational complexity. The proposed technique resulted in PSNR and SSIM values were ~3%-16% higher than any of the other traditional algorithms tested, and our method recovered more clear textures from noisy images compared with deep-learning-based methods. As such, the presented technique is a viable new approach to MR upsampling, particularly for noisy images.

Defects Detection of Digital Radiographic Images of Aircraft Structure Materials via Geometric Locally Adaptive Sharpening

ABSTRACTThe life of an aircraft depends on the early detection and removal of corrosion in its structure. The importance of detecting corrosion cannot be understated, because corrosion can cause other kinds of damage, such as cracks. Radiography is an important method for the detection of hidden defects in aircraft structure. To maximize information extraction from the radiographic images, the noise of the system should be minimized, or the contrast of the defective region should be maximized by different methods. The development of effective image processing methods, within both the spatial and frequency domains, is important to the research of industrial radiographic testing. In this study, the geometric locally adaptive sharpening method was used to improve hidden structure visualization of details and defects from aircraft part radiographs. The method relies on sharpening by using the steering kernel regression method. Here, the enhancing contrast and sharpening algorithm are effectively mixed together. The proposed algorithm was successfully applied to radiographic images of aircraft parts. An improvement of the structure detail visualization and defect region detection was achieved by sharpening the edges and preserving fine detail imaging information. Experts’ reviews showed that defect regions from the geometric locally adaptive sharpening reconstructed images were better visualized than the original images. Also, the resulting evaluation of the output images shows that the edges are sharpened by the proposed method and that the background of the image decreases to zero.

Detail Preserved Single Image Dehazing Algorithm Based on Airlight Refinement

Single-image haze removal is important for many practical applications (e.g., surveillance). However, dehazed results of existing algorithms tend to be oversmoothed with missing fine image details. This drawback is caused by two factors: inaccurate airlight estimations and disregarding multiple scattering. In this paper, we propose a detail-preserving image dehazing algorithm based on two key priors, namely, the depth-edge aware prior and the airlight impact regularity prior. The proposed algorithm makes contributions in both the haze removal step and the postprocessing step. First, based on the depth-edge aware prior, an airlight refinement algorithm is proposed. The gradient strength of the minimum channel is employed to calculate punishment weights to smooth the dark channel. Second, based on the airlight impact regularity prior, an adaptive sharpening model that considers the refined airlight to determine the sharpening strength value is established to enhance levels of detail. Experimental results demonstrate that the proposed algorithm cannot only effectively remove haze but can also enhance levels of detail to thus outperform the state of the art on a wide variety of images.

Context-dependent coding and gain control in the auditory system of crickets.

Sensory systems process stimuli that greatly vary in intensity and complexity. To maintain efficient information transmission, neural systems need to adjust their properties to these different sensory contexts, yielding adaptive or stimulus-dependent codes. Here, we demonstrated adaptive spectrotemporal tuning in a small neural network, i.e. the peripheral auditory system of the cricket. We found that tuning of cricket auditory neurons was sharper for complex multi-band than for simple single-band stimuli. Information theoretical considerations revealed that this sharpening improved information transmission by separating the neural representations of individual stimulus components. A network model inspired by the structure of the cricket auditory system suggested two putative mechanisms underlying this adaptive tuning: a saturating peripheral nonlinearity could change the spectral tuning, whereas broad feed-forward inhibition was able to reproduce the observed adaptive sharpening of temporal tuning. Our study revealed a surprisingly dynamic code usually found in more complex nervous systems and suggested that stimulus-dependent codes could be implemented using common neural computations.

Super Resolution Reconstruction in Mixed Noise Environment

A hybrid Super Resolution (SR) algorithm is proposed to deal with the Low Resolution (LR) images degraded by Mixed (Gaussian + Impulse) noise. The algorithm adaptively estimates and removes the impulse noise from the input LR images based on edge, geometrical & size characteristics. The fuzzy based impulse noise removal algorithm is along with adaptive sharpening filter based SR using steering kernel regression are used to obtain a HR image. The experimental results confirm the efficacy of the algorithm for different types of images at various noise densities. General Terms Super Resolution, Mixed Noise, Surveillance applications,

Noise-robust video super-resolution using an adaptive spatial-temporal filter

In this paper, we introduce a new interpolation-based super-resolution scheme for super-resolving a low-resolution video that contains large-scale local motions and/or heavy noise. Our scheme leverages an efficient space-time descriptor to adapt the interpolation kernel to the video's spatial and temporal structures. Nevertheless, in the presence of large-scale local motions, the kernel suffers from tracking the motions incorrectly, leading to inaccurate temporal averaging. To address this problem, prior to computing the interpolation kernel, a mobile-neighborhood strategy that can identify the appropriate neighborhoods in adjacent frames is applied to neutralize the large-scale motions. Furthermore, we incorporate an adaptive sharpening technique into the kernel computation to remove the background noise and enhance the fine details simultaneously. Extensive experimental results on real-world videos show that the proposed method outperforms certain other state-of-the-art video super-resolution algorithms both visually and quantitatively, particularly in the presence of large-scale motions and/or heavy noise.

A new method for fusion, denoising and enhancement of x-ray images retrieved from Talbot–Lau grating interferometry

This paper introduces a new image denoising, fusion and enhancement framework for combining and optimal visualization of x-ray attenuation contrast (AC), differential phase contrast (DPC) and dark-field contrast (DFC) images retrieved from x-ray Talbot–Lau grating interferometry. The new image fusion framework comprises three steps: (i) denoising each input image (AC, DPC and DFC) through adaptive Wiener filtering, (ii) performing a two-step image fusion process based on the shift-invariant wavelet transform, i.e. first fusing the AC with the DPC image and then fusing the resulting image with the DFC image, and finally (iii) enhancing the fused image to obtain a final image using adaptive histogram equalization, adaptive sharpening and contrast optimization. Application examples are presented for two biological objects (a human tooth and a cherry) and the proposed method is compared to two recently published AC/DPC/DFC image processing techniques. In conclusion, the new framework for the processing of AC, DPC and DFC allows the most relevant features of all three images to be combined in one image while reducing the noise and enhancing adaptively the relevant image features. The newly developed framework may be used in technical and medical applications.

An Improved Low-cost Adaptive Bicubic Interpolation Arithmetic and VLSI Implementation

A novel scaling algorithm is proposed which consists of a bicubic interpolation and an adaptive sharpening filter. The proposed sharpening filter is added to mitigate the blurring effects existing in bicubic interpolation methods. We also verify the scaling quality by taking into account the adaptive technique. Furthermore, we present both the procedures of filtering before and after interpolation in order to reduce the overall computing time. Compared with the previous reported techniques, our method performs better in terms of both quantitative evaluation and visual quality. To achieve the goal of real time and low cost, we describe a pipelined very large scale integration (VLSI) architecture for the implementation of the algorithm. The VLSI architecture of our image scaling processor contains 695 LEs and yields a processing rate of about 165 MHz by using field-programmable gate array (FPGA) technology. Our proposed architecture reduces the amount of gates by 36.8% while achieves an average PSNR increase of 1.5 dB in image quality.

Adaptive sharpening of depth maps for 3D-TV

Depth maps, which can be represented as greyscale images, are used to aid rendering of novel views in three-dimensional (3D) video systems. However, compressing them using existing video codecs, such as H.264/AVC, leads to low quality rendered views. Presented is a sharpening method based on adaptive bilateral filtering to eliminate certain artifacts observed in compressed depth maps to improve the quality of rendered views. Experimental results demonstrate that significant rendering quality improvements of up to 1.9 dB can be achieved with the proposed method.

Masking fields: a massively parallel neural architecture for learning, recognizing, and predicting multiple groupings of patterned data

A massively parallel neural network architecture, called a masking field, is characterized through systematic computer simulations. A masking field is a multiple-scale self-similar automatically gain-controlled cooperative- competitive feedback network F(2). Network F(2) receives input patterns from an adaptive filter F(1) ? F(2) that is activated by a prior processing level F(1). Such a network F(2) behaves like a content-addressable memory. It activates compressed recognition codes that are predictive with respect to the activation patterns flickering across the feature detectors of F(1) and competitively inhibits, or masks, codes which are unpredictive with respect to the F(1) patterns. In particular, a masking field can simultaneously detect multiple groupings within its input patterns and assign activation weights to the codes for these groupings which are predictive with respect to the contextual information embedded within the patterns and the prior learning of the system. A masking field automatically rescales its sensitivity as the overall size of an input pattern changes, yet also remains sensitive to the microstructure within each input pattern. In this way, a masking field can more strongly activate a code for the whole F(1) pattern than for its salient parts, yet amplifies the code for a pattern part when it becomes a pattern whole in a new input context. A masking field can also be primed by inputs from F(1): it can activate codes which represent predictions of how the F(1) pattern may evolve in the subsequent time interval. Network F(2) can also exhibit an adaptive sharpening property: repetition of a familiar F(1) pattern can tune the adaptive filter to elicit a more focal spatial activation of its F(2) recognition code than does an unfamiliar input pattern. The F(2) recognition code also becomes less distributed when an input pattern contains more contextual information on which to base an unambiguous prediction of which the F(1) pattern is being processed. Thus a masking field suggests a solution of the credit assignment problem by embodying a real-time code for the predictive evidence contained within its input patterns. Such capabilities are useful in speech recognition, visual object recognition, and cognitive information processing. An absolutely stable design for a masking field is disclosed through an analysis of the computer simulations. This design suggests how associative mechanisms, cooperative-competitive interactions, and modulatory gating signals can be joined together to regulate the learning of compressed recognition codes. Data about the neural substrates of learning and memory are compared to these mechanisms.