Deploying deep neural networks (DNNs) in resource-limited environments—such as smartwatches, IoT nodes, and intelligent sensors—poses significant challenges due to constraints in memory, computing power, and energy budgets. This paper presents a comprehensive review of recent advances in accelerating DNN inference on edge platforms, with a focus on model compression, compiler optimizations, and hardware–software co-design. We analyze the trade-offs between latency, energy, and accuracy across various techniques, highlighting practical deployment strategies on real-world devices. In particular, we categorize existing frameworks based on their architectural targets and adaptation mechanisms and discuss open challenges such as runtime adaptability and hardware-aware scheduling. This review aims to guide the development of efficient and scalable edge intelligence solutions.

This paper proposes a single image haze removal algorithm that shows a marked improvement on the color attenuation prior-based method. Through a vast number of experiments on a wide variety of images, it is discovered that there are problems in the color attenuation prior, such as color distortion and background noise, which arise due to the fact that the priors do not hold true in all circumstances. Successful resolution of these problems using the proposed algorithm shows its superior performance to other state-of-the-art methods in terms of both subjective visual quality and quantitative metrics, on both synthetic and natural hazy image datasets. The proposed algorithm also is computationally friendly, due to the use of an efficient quad-decomposition algorithm for atmospheric light estimation and a simple modified hybrid median filter for depth map refinement.

Vision-based systems operating outdoors are significantly affected by weather conditions, notably those related to atmospheric turbidity. Accordingly, haze removal algorithms, actively being researched over the last decade, have come into use as a pre-processing step. Although numerous approaches have existed previously, an efficient method coupled with fast implementation is still in great demand. This paper proposes a single image haze removal algorithm with a corresponding hardware implementation for facilitating real-time processing. Contrary to methods that invert the physical model describing the formation of hazy images, the proposed approach mainly exploits computationally efficient image processing techniques such as detail enhancement, multiple-exposure image fusion, and adaptive tone remapping. Therefore, it possesses low computational complexity while achieving good performance compared to other state-of-the-art methods. Moreover, the low computational cost also brings about a compact hardware implementation capable of handling high-quality videos at an acceptable rate, that is, greater than 25 frames per second, as verified with a Field Programmable Gate Array chip. The software source code and datasets are available online for public use.

Haze removal is an ill-posed problem that has attracted much scientific interest due to its various practical applications. Existing methods are usually founded upon various priors; consequently, they demonstrate poor performance in circumstances in which the priors do not hold. By examining hazy and haze-free images, we determined that haze density is highly correlated with image features such as contrast energy, entropy, and sharpness. Then, we proposed an iterative algorithm to accurately estimate the extinction coefficient of the transmission medium via direct optimization of the objective function taking into account all of the features. Furthermore, to address the heterogeneity of the lightness, we devised adaptive atmospheric light to replace the homogeneous light generally used in haze removal. A comparative evaluation against other state-of-the-art approaches demonstrated the superiority of the proposed method. The source code and data sets used in this paper are made publicly available to facilitate further research.

Haze is a term that is widely used in image processing to refer to natural and human-activity-emitted aerosols. It causes light scattering and absorption, which reduce the visibility of captured images. This reduction hinders the proper operation of many photographic and computer-vision applications, such as object recognition/localization. Accordingly, haze removal, which is also known as image dehazing or defogging, is an apposite solution. However, existing dehazing algorithms unconditionally remove haze, even when haze occurs occasionally. Therefore, an approach for haze density estimation is highly demanded. This paper then proposes a model that is known as the haziness degree evaluator to predict haze density from a single image without reference to a corresponding haze-free image, an existing georeferenced digital terrain model, or training on a significant amount of data. The proposed model quantifies haze density by optimizing an objective function comprising three haze-relevant features that result from correlation and computation analysis. This objective function is formulated to maximize the image’s saturation, brightness, and sharpness while minimizing the dark channel. Additionally, this study describes three applications of the proposed model in hazy/haze-free image classification, dehazing performance assessment, and single image dehazing. Extensive experiments on both real and synthetic datasets demonstrate its efficacy in these applications.

In recent years, machine vision algorithms have played an influential role as core technologies in several practical applications, such as surveillance, autonomous driving, and object recognition/localization. However, as almost all such algorithms are applicable to clear weather conditions, their performance is severely affected by any atmospheric turbidity. Several image visibility restoration algorithms have been proposed to address this issue, and they have proven to be a highly efficient solution. This paper proposes a novel method to recover clear images from degraded ones. To this end, the proposed algorithm uses a supervised machine learning-based technique to estimate the pixel-wise extinction coefficients of the transmission medium and a novel compensation scheme to rectify the post-dehazing false enlargement of white objects. Also, a corresponding hardware accelerator implemented on a Field Programmable Gate Array chip is in order for facilitating real-time processing, a critical requirement of practical camera-based systems. Experimental results on both synthetic and real image datasets verified the proposed method’s superiority over existing benchmark approaches. Furthermore, the hardware synthesis results revealed that the accelerator exhibits a processing rate of nearly 271.67 Mpixel/s, enabling it to process 4K videos at 30.7 frames per second in real time.