scholarly journals Enhanced Tone Mapping Using Regional Fused GAN Training with a Gamma-Shift Dataset

2021 ◽  
Vol 11 (16) ◽  
pp. 7754
Author(s):  
Sung-Woon Jung ◽  
Hyuk-Ju Kwon ◽  
Sung-Hak Lee

High-dynamic-range (HDR) imaging is a digital image processing technique that enhances an image’s visibility by modifying its color and contrast ranges. Generative adversarial networks (GANs) have proven to be potent deep learning models for HDR imaging. However, obtaining a sufficient volume of training image pairs is difficult. This problem has been solved using CycleGAN, but the result of the use of CycleGAN for converting a low-dynamic-range (LDR) image to an HDR image exhibits problematic color distortion, and the intensity of the output image only slightly changes. Therefore, we propose a GAN training optimization model for converting LDR images into HDR images. First, a gamma shift method is proposed for training the GAN model with an extended luminance range. Next, a weighted loss map trains the GAN model for tone compression in the local area of images. Then, a regional fusion training model is used to balance the training method with the regional weight map and the restoring speed of local tone training. Finally, because the generated module tends to show a good performance in bright images, mean gamma tuning is used to evaluate image luminance channels, which are then fed into modules. Tests are conducted on foggy, dark surrounding, bright surrounding, and high-contrast images. The proposed model outperforms conventional models in a comparison test. The proposed model complements the performance of an object detection model even in a real night environment. The model can be used in commercial closed-circuit television surveillance systems and the security industry.

2015 ◽  
Vol 16 (1) ◽  
pp. 136
Author(s):  
Behrouz Memarzadeh ◽  
Mohammad Ali Mohammadi

Vision-based flame detection has drawn significant attention in the past decade with camera surveillance systems becoming ubiquitous. This paper proposes a multi criterion method to detect fire or flames by processing the video data generated by a high speed camera. Since flame images are special class of images, some of the unique features of a flame may be used to identify flame. There are some differences between flame images and other general images. By using these features we are able to detect fire correctly with least false alarm. In this paper we present an algorithm which can detect fire and reduce number of false alarms by counting number of identified pixels. In the algorithm, we preprocess the images to have better results. So first we adjust the gray level of a flame image according to its statistical distribution to have better processing. After that we try to extract fire features in images. First by using color characteristics, the ratio of red to green, we can identify probable fire-like or fire like pixels. Second, to highlight the regions with high gray level contrast at their edges, we use the extended prewitt filter. We use AND operation on two above processing images to remove unrelated pixels, at last by using flicker frequency, the oscillating change in the number of identified pixels over time is transformed into the frequency domain to complete detection algorithm. Simulation proves the algorithm ability to detect fire in different situations in video sequences.


This paper depicts the realization of DIP (Digital Image Processing) technique for pattern recognition to identify objects in video stream. The proposed model compares the test object with standard model and identifies the missing objects in the test item. The model uses image classifier algorithm as a tool. The simulations are carried out in MATLab Simulink and various test items are compared under different morphological conditions. The model is fabricated to analyze and indicate the omitted components in wind turbine.


2020 ◽  
Vol 2020 (7) ◽  
pp. 213-1-213-7
Author(s):  
Nabeel. A. Riza ◽  
Nazim Ashraf

Proposed for the first time is a novel calibration empowered minimalistic multi-exposure image processing technique using measured sensor pixel voltage output and exposure time factor limits for robust camera linear dynamic range extension. The technique exploits the best linear response region of an overall nonlinear response image sensor to robustly recover via minimal count multi-exposure image fusion, the true and precise scaled High Dynamic Range (HDR) irradiance map. CMOS sensor-based experiments using a measured Low Dynamic Range (LDR) 44 dB linear region for the technique with a minimum of 2 multi-exposure images provides robust recovery of 78 dB HDR low contrast highly calibrated test targets.


Author(s):  
GUOJUN LIU ◽  
XIANGCHU FENG ◽  
WEIWEI WANG ◽  
XUANDE ZHANG

Wavelet has become an appealing image processing technique, due to the fact that the sparseness of wavelet expansion is equivalent to smoothness measure in Besov spaces so that the regularization of image can be performed by manipulating its wavelet coefficients. Unfortunately, wavelets have good performance especially at representing point singularities, but they fail to efficiently represent object edges. As one of computational harmonic analysis tools, curvelets have an essentially optimal representation of objects which is C2 away from a C2 edge. In this paper, we first apply constraint of curvelet-type decomposition spaces as a regularizing term to variational model for image denoising. Based on the equivalent relationship between semi-norm of curvelet-type decomposition spaces and the weighted curvelet coefficients, solution to the proposed model approximately equals to different curvelet shrinkages. As a second contribution, we also propose another image restoration model from image decomposition point of view. Furthermore, an equivalent theorem of two proposed models is given. Finally, the experiment results show the superiority of proposed models over traditional wavelet-based ones.


2020 ◽  
Vol 2 (1) ◽  
pp. 55-64 ◽  
Author(s):  
Mr. H. James Deva Koresh

Human eye always reveals a non –linear understanding, for the disturbances caused by the lossy image and video coding. This is mainly because of the masking capability of the human eye to conceal the attributes such as contrast, luminous, spatial and temporal frequencies. To have a distortion less and efficient video encoding for the high dynamic video range content by eluding the invisible messages in the video that causes disturbances the paper puts forth the quantization with perception utilizing the luminous masking. The methodology utilized, computes the tone mapping to scale every frames in the HDR and later quantizes on unit basis with perception tuning. For this purpose the mechanism put forth incorporates the reference model of the HEVC with the extension range of the HEVC. The proposed model was validated by evaluating the reduction incurred in each rate of bit compared to the HDR range extension. The results acquired proved to have an enhancement in terms of the savings endured in the bit rate compared to the High efficient video coding that relied on the high dynamic range visible difference predictor-II


2021 ◽  
Vol 18 (2) ◽  
pp. 379-399 ◽  
Author(s):  
Jaepil Ko ◽  
Kyung Cheoi

In this paper, a hybrid visual attention model to effectively detect a distant target is proposed. The model employs the human visual attention mechanism and consists of two models, the training model, and the detection model. In the training model, some of the features are selected to train in the process of extracting and combining the early visual features from the training image of the target by bottom-up manner, and these features are trained and accumulated as trained data. When the image containing the target is input into the detection model, a task of selectively promoting only features of the target using pre-trained data is performed. As a result, the desired target is detected through the saliency map created as a result of the feature combination. The model has been tested on various images, and the experimental results demonstrate that the proposed model detected the target more accurately and faster than other previous models.


2019 ◽  
Vol 13 (2) ◽  
pp. 132
Author(s):  
Sumaia Saraireh ◽  
Ahmad Hassanat ◽  
Mohammad Abu Al-Taieb ◽  
Hashem A Kilani

This work provides a new dataset method intended to build a biomechanical training model for the free-throws shots in basketball. Eight youth players from Jordanian secondary public school were video recorded from the sagittal plane executing free throw shots in basketball. Collectively (480) video clips were recorded and analyzed using image processing techniques to identify the ball track. Video processing involves extracting (11) different parameters that may affect the free throw in basketball game after detecting the ball trajectory. Creation of this dataset and its subsequent use for extracting free-throws information yielded several insights. First, a set of most important features were identified as those affecting the free-throws score in basketball. Second, our data set can be trained and tested using machine learning classifiers for building a new biomechanical training model based on set of rules that can be useful for both trainers and trainee to rehearse on successful free-throws in basketball. The dataset is being made publicly available at www.ju.edu.jo.


Sign in / Sign up

Export Citation Format

Share Document