In this paper, we propose a retrieval method based on Multimodal Pedestrian Features. This method aims to combine pedestrian attribute features, as described in text, with video images to automatically search for and ...
详细信息
images and videos captured in poor illumination conditions are degraded by low brightness, reduced contrast, color distortion, and noise, rendering them barely discernable for human perception and ultimately negativel...
详细信息
ISBN:
(纸本)9781510673854;9781510673847
images and videos captured in poor illumination conditions are degraded by low brightness, reduced contrast, color distortion, and noise, rendering them barely discernable for human perception and ultimately negatively impacting computer vision system performance. These challenges are exasperated when processing video surveillance camera footage, using this unprocessed video data as-is for real-time computer vision tasks across varying environmental conditions within Intelligent Transportation systems (ITS), such as vehicle detection, tracking, and timely incident detection. The inadequate performance of these algorithms in real-world deployments incurs significant operational costs. Low-light image enhancement (LLIE) aims to improve the quality of images captured in these unideal conditions. Groundbreaking advancements in LLIE have been recorded employing deep-learning techniques to address these challenges, however, the plethora of models and approaches is varied and disparate. This paper presents an exhaustive survey to explore a methodical taxonomy of state-of-the-art deep learning-based LLIE algorithms and their impact when used in tandem with other computer vision algorithms, particularly detection algorithms. To thoroughly evaluate these LLIE models, a subset of the BDD100K dataset, a diverse real-world driving dataset is used for suitable image quality assessment and evaluation metrics. This study aims to provide a detailed understanding of the dynamics between low-light image enhancement and ITS performance, offering insights into both the technological advancements in LLIE and their practical implications in real-world conditions. The project Github repository can be accessed here.
High-frequency imaging sonar systems are a critical technology for underwater sensing. Design and experimental validation of signal processingalgorithms for image reconstruction is very challenging due to difficultie...
详细信息
ISBN:
(纸本)9798350362077
High-frequency imaging sonar systems are a critical technology for underwater sensing. Design and experimental validation of signal processingalgorithms for image reconstruction is very challenging due to difficulties in deployment and accurate positioning. Modelling and simulation are powerful tools to evaluate the system before prototyping. Unfortunately, there are only a few simulation tools available, and the computation complexity for accurate simulation is very high. Most of the simulators directly convert the estimated target range, receive beam direction and reflected intensity level in to synthetic sonar image. These simulators are not accounting for the spatial response of the array with beamformer and temporal response of the waveform, while generating the image. To evaluate the effectiveness of different signal processing techniques for image reconstruction, a received signal model for the sensor array considering the array and waveform properties is essential. This paper presents an efficient signal and noise simulator, to generate the sonar receiver array signal time series for practical imaging scenarios, with five times lesser computation time compared to the state-of-the-art simulator reported in the literature. The simulator also contains a basic sonar processing module which performs near-field beamforming, matched filtering and imageprocessing on the simulated signal, to generate the final sonar image. Optimisation in the sonar receiver and signal processing can be conveniently evaluated through this simulation scheme. The simulation results are compared with experimental results to verify the simulation model. Being generic in nature this simulator can simulate sector scan sonar, single beam and multi-beam side scan sonars simultaneously for the same simulation scenario. This enables verification of co-registration of side scan and gap filler sonar images.
Intelligent control and computer vision algorithms can achieve clearer, more delicate, and realistic image effects through extremely high image quality and target recognition technology, as well as complete image proc...
详细信息
In the era of big data, imageprocessing still faces significant bottlenecks compared to other fields of computer science. In this paper we studied the feature point extraction algorithm based on gray points and the s...
详细信息
The recently introduced Consistency models pose an efficient alternative to diffusion algorithms, enabling rapid and good quality image synthesis. These methods overcome the slowness of diffusion models by directly ma...
This research presents an innovative approach to accurately extract multi-scale landscape characters by combining geographic information systems (GIS) and graphic processingalgorithms. It is essential to recognize an...
详细信息
Skin cancer, which primarily impacts skin exposed to ultraviolet (UV) rays against the sun, represents dangerous to the most significant organs in the human body, the skin. Usually, a spot, lump, or mole that appears ...
详细信息
ISBN:
(纸本)9798350377972
Skin cancer, which primarily impacts skin exposed to ultraviolet (UV) rays against the sun, represents dangerous to the most significant organs in the human body, the skin. Usually, a spot, lump, or mole that appears on the skin is the primary suspicion of skin cancer. However, each of these can undergo changes in coloring or shape as time passes. Recovery for skin cancer is mostly possible if the disease is discovered early. Numerous medical diagnostic methods, such as Dermoscopy, biopsy, and ocular examination of the affected area, are useful in helping anticipate the development of skin cancer. However, these approaches have the disadvantage of delivering erroneous results because it is extremely difficult to distinguish between normal and malignant skin. Therefore, the drawback of these diagnostic procedures is that machine learning algorithms are currently used together with imageprocessing techniques to examine the images for the purpose of precisely identifying skin cancer. The current research employs the ISIC dataset to develop a novel model for skin cancer classification that combines imageprocessing techniques with advanced machine learning methods, including Crammer-Singer Support vector machine learning algorithms. The categorization of skin cancer begins with preprocessing the input image, which includes hair removal using a morphological filter and image enhancement using a median filter to minimize noise and increase image clarity. The ABCD approach is used to segment lesion images by evaluating them for asymmetry, border irregularity, color variability, and diameter. The crammer-Singer SVM algorithm is then used with these images to classify skin lesions into various types such as melanoma (MEL), melanocytic nevus (NV), basal cell carcinoma (BCC), actinic keratosis (AK), benign keratosis (BKL), dermatofibroma (DF), vascular lesion (VASC), and squamous cell carcinoma (SCC), leveraging its robust multi-class handling capabilities. The system achieve
With the continuous advancement of smart city construction, autonomous driving technology is playing an increasingly important role in urban traffic systems. This study aims to explore the development and optimization...
详细信息
Deep Learning is a technology developed with GPU Acceleration that has a good ability to process image computations. One of the deep learning methods widely used in classifying two-dimensional objects is the Convoluti...
详细信息
ISBN:
(纸本)9783031214370;9783031214387
Deep Learning is a technology developed with GPU Acceleration that has a good ability to process image computations. One of the deep learning methods widely used in classifying two-dimensional objects is the Convolutional Neural Network algorithm. Just like other imageprocessingalgorithms, the classification process is very dependent on the quality of the image used. Therefore, it is concerned that pre-processing is done. This study aimed to find a scenario for image data pre-processing by comparing the threshold types used. By using two scenarios, the first scenario using Simple Threshold and the second scenario using threshold Canny. The first scenario begins with collecting data from an X-ray image after the established dataset is advanced to pre-process the data set. In this pre-processing data, several things were done to increase the level of data accuracy by changing and equalizing the pixel size in the dataset, changing the color of the image on the dataset to grayscale, distributing the histogram or commonly known as histogram equalization, and finally applying a simple threshold. Unlike the second scenario, which does not use a simple threshold but uses a threshold canny. After completion of the pre-processing stage and then the continued training phase. At this stage, the dataset will be trained using CNN. After the dataset is trained, it enters the testing stage. The testing stage shows the results that the data is classified properly. The validation obtained from the two scenarios shows that the simple threshold gives better results than the canny threshold, with a value that shows a simple threshold of 97% and a canny threshold of 89%. This result shows that the dataset's treatment differences greatly affect the results' accuracy.
暂无评论