Dim infrared image enhancement based on convolutional neural network

Fan, Zunlin; Du-yan, BI; Xiong, Lei; Ma, Shiping; He, Linyuan; Ding, Wenshan

doi:10.1016/j.neucom.2017.07.017

Cited by 77 publications

(36 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Wang et al [24] use the open dataset (ILSVRC 2013) and the simulated dataset to train data and extract features of infrared small targets. Fan et al [25] employ the MNIST dataset to predict the dim targets and background sub-images in the first layer of CNN. Liu et al [26] also use visible images (Ima-geNet dataset) to train a CNN, and then parameters are adjusted to deal with small target detection and tracking in infrared images.…”

Section: B Multi-frame Detection Methodsmentioning

confidence: 99%

“…Liu et al [26] also use visible images (Ima-geNet dataset) to train a CNN, and then parameters are adjusted to deal with small target detection and tracking in infrared images. On the whole, due to the lack of sufficient training data, these methods have to introduce visible datasets [25], [26] or simulated datasets [24], [45], [46] to train a CNN. However, when an infrared small target appears in a low local contrast image or the size of an infrared target is too small, it is still weak for CNN on feature learning.…”

Section: B Multi-frame Detection Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Infrared Small Target Detection Using a Temporal Variance and Spatial Patch Contrast Filter

Gao

Lin

2019

IEEE Access

View full text Add to dashboard Cite

Infrared small target detection is challenging due to the various background and low signalto-clutter ratios. Considering the information deficiency faced by single spatial or temporal information, we construct a low false alarm spatial and temporal filter for infrared small target detection. A multiscale patch-based contrast measure is first used to suppress background and remove cloud edges at a coarse level. Then, a temporal variance filter is used to remove small broken cloud regions and suppress noise at a fine level. By integrating these two methods, infrared small targets can be extracted accurately and robustly using an adaptive threshold segmentation. The experimental results indicate that our proposed method can robustly detect small infrared targets with a low false alarm rate. INDEX TERMS Infrared image sequence, small target detection, local contrast, temporal profile.

show abstract

Section: B Multi-frame Detection Methodsmentioning

confidence: 99%

Section: B Multi-frame Detection Methodsmentioning

confidence: 99%

Infrared Small Target Detection Using a Temporal Variance and Spatial Patch Contrast Filter

Gao

Lin

2019

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Recently, several convolution neural network-(CNN-) based infrared small target detection methods have been proposed [27][28][29][30]. Fan et al [27] simulated the training data using the MNIST database (http://yann.lecun.com/exdb/ mnist/) containing handwritten images and labels, and then the target and background are estimated based on the convolution kernels in the first layer of the CNN architecture, which are learned from the input data. e main disadvantage of applying the deep learning method to infrared small target detection is that feature learning will be very difficult for the CNN-based method as the infrared small target has no remarkable texture and shape features.…”

Section: Introductionmentioning

confidence: 99%

“…e main disadvantage of applying the deep learning method to infrared small target detection is that feature learning will be very difficult for the CNN-based method as the infrared small target has no remarkable texture and shape features. In addition, because of insufficient training data, the visible datasets [27] or simulated infrared datasets [29,30] are used to train the detection model. As a result, generalizability of the detection method is also limited.…”

Section: Introductionmentioning

confidence: 99%

Infrared Small Target Detection with Total Variation and Reweighted ℓ1 Regularization

Fang

Chen

Liu

et al. 2020

Mathematical Problems in Engineering

View full text Add to dashboard Cite

Infrared small target detection plays an important role in infrared search and tracking systems applications. It is difficult to perform target detection when only a single image with complex background clutters and noise is available, where the key is to suppress the complex background clutters and noise while enhancing the small target. In this paper, we propose a novel model for separating the background from the small target based on nonlocal self-similarity for infrared patch-image. A total variation-based regularization term for the small target image is incorporated into the model to suppress the residual background clutters and noise while enhancing the smoothness of the solution. Furthermore, a reweighted sparse constraint is imposed for the small target image to remove the nontarget points while better highlighting the small target. For higher computational efficiency, an adapted version of the alternating direction method of multipliers is employed to solve the resulting minimization problem. Comparative experiments with synthetic and real data demonstrate that the proposed method is superior in detection performance to the state-of-the-art methods in terms of both objective measure and visual quality.

show abstract

“…This algorithm made the detection accuracy up to 93% [8]. Fan et al [9] proposed an infrared image enhancement method based on convolutional neural network, which could enhance the infrared target area in the state of background clutters and low contrast [10]. In order to design a deep network with better self-learning ability and self-adaptive ability, the temporal and spatial features of motion imagery were used in artificial neural network (ANN) for infrared small target detection [11].…”

Section: Introductionmentioning

confidence: 99%

Infrared Target Extraction Based on Immune Extension Neural Network

Dong

Hao

et al. 2020

IEEE Access

View full text Add to dashboard Cite

This paper addresses the difficulties of target extraction from low contrast low resolution infrared images. The extension neural network is integrated with the negative selection algorithm (NSA) in artificial immunity. Based on the specific recognition principle of long non-coding RNA (lncRNA) in biological immune system, the mature self-extension detectors are calculated to extract infrared target and background. Upon the convergence of clustering process, the output layer of network merges the categories corresponding to the target to be detected. The proposed method, it needs less detectors than the traditional NSA algorithm. Compared with several target detection algorithms, the proposed algorithm has higher detection accuracy and higher time efficiency. INDEX TERMS Artificial immunity, extension analysis, negative selection algorithm (NSA), long noncoding RNA (lncRNA), infrared target extraction.

show abstract

Dim infrared image enhancement based on convolutional neural network

Cited by 77 publications

References 30 publications

Infrared Small Target Detection Using a Temporal Variance and Spatial Patch Contrast Filter

Infrared Small Target Detection Using a Temporal Variance and Spatial Patch Contrast Filter

Infrared Small Target Detection with Total Variation and Reweighted ℓ1 Regularization

Infrared Target Extraction Based on Immune Extension Neural Network

Contact Info

Product

Resources

About