The non-uniformity of the readout circuit response in the infrared focal plane array unit detector can result in fixed pattern noise with stripe, which seriously affects the quality of the infrared images. Considering the problems of existing non-uniformity correction, such as the loss of image detail and edge blurring, a multi-scale residual network with attention mechanism is proposed for single infrared image stripe noise removal. A multi-scale feature representation module is designed to decompose the original image into varying scales to obtain more image information. The product of the direction structure similarity parameter and the Gaussian weighted Mahalanobis distance is used as the similarity metric; a channel spatial attention mechanism based on similarity (CSAS) ensures the extraction of a more discriminative channel and spatial feature. The method is employed to eliminate the stripe noise in the vertical and horizontal directions, respectively, while preserving the edge texture information of the image. The experimental results show that the proposed method outperforms four state-of-the-art methods by a large margin in terms of the qualitative and quantitative assessments. One hundred infrared images with different simulated noise intensities are applied to verify the performance of our method, and the result shows that the average peak signal-to-noise ratio and average structural similarity of the corrected image exceed 40.08 dB and 0.98, respectively.

We propose a sub-wavelength range-based dual-band tunable ideal terahertz metamaterial perfect absorber. The absorber structure consists of three main layers, with the absorber layer consisting of a metal I-shaped structure. By simulating the incident wave absorbance of the structure, we found that the structure has more than 99% absorption peaks in both bands. In addition, we have investigated the relationship between structural absorbance and the structural geometrical parameters. We have studied the relationship between the thickness of the metal absorber layer $h_b$ and the absorbance of the metamaterial structure in the 4–14 THz band. Secondly, we have studied the relationship between the thickness of the SiO₂ dielectric layer and structural absorbance. Afterwards, we have studied the relationship between the incident angle of the incident electromagnetic wave and structural absorbance. Finally, we have studied the relationship between the length of the metal structure and structural absorbance. The structure can be effectively used for detectors, thermal emitters, terahertz imaging and detection.

Blood samples are easily damaged in traditional bloodstain detection and identification. In complex scenes with interfering objects, bloodstain identification may be inaccurate, with low detection rates and false-positive results. In order to meet these challenges, we propose a bloodstain detection and identification method based on hyperspectral imaging and mixed convolutional neural networks, which enables fast and efficient non-destructive identification of bloodstains. In this study, we apply visible/near-infrared reflectance hyperspectral imaging in the 380-1000 nm spectral region to analyze the shape, structure, and biochemical characteristics of bloodstains. Hyperspectral images of bloodstains on different substrates and six bloodstain analogs are experimentally obtained. The acquired spectral pixels are pre-processed by Principal Component Analysis (PCA). For bloodstains and different bloodstain analogs, regions of interest are selected from each substance to obtain pixels, which are further used in convolutional neural network (CNN) modeling. After the mixed CNN modeling is completed, pixels are selected from the hyperspectral images as a test set for bloodstains and bloodstain analogs. Finally, the bloodstain recognition ability of the mixed 2D-3D CNN model is evaluated by analyzing the kappa coefficient and classification accuracy. The experimental results show that the accuracy of the constructed CNN bloodstain identification model reaches 95.4%. Compared with other methods, the bloodstain identification method proposed in this study has higher efficiency and accuracy in complex scenes. The results of this study will provide a reference for the future development of the bloodstain online detection system.

The dynamic range of an X-ray digital imaging system is very important when detecting objects with a high absorption ratio. In this paper, a ray source filter is used to filter the low-energy ray components which have no penetrating power to the high absorptivity object to reduce the X-ray integral intensity. This enables the effective imaging of the high absorptivity objects and avoids the image saturation of low absorptivity objects, thus achieving single exposure imaging of high absorption ratio objects. However, this method will reduce the image contrast and weaken the image structure information. Therefore, this paper proposes a contrast enhancement method for X-ray images based on Retinex. Firstly, based on Retinex theory, the multi-scale residual decomposition network decomposes the image into an illumination component and a reflection component. Then, the contrast of the illumination component is enhanced through the U-Net model with the global–local attention mechanism, and the reflection component is enhanced in detail using the anisotropic diffused residual dense network. Finally, the enhanced illumination component and the reflected component are fused. The results show that the proposed method can effectively enhance the contrast in X-ray single exposure images of the high absorption ratio objects, and can fully display the structure information of images on devices with low dynamic range.

Low light level (LLL) images, which were captured by Intensified CCD (ICCD) camera equipped with an image intensifier, suffer low spatial resolution and contrast due to noise and dispersion. By dividing the integration time into intervals short enough, we obtain photon images where photon formed spots were nearly nonoverlapping. In order to enhance LLL images, we propose a Multi-layer slicing (MLS) photon localization algorithm based on photon images. The photon image is sliced by different planes. Photon spatial distribution (PSD) information is acquired by using projected area ratio, circularity and the number of slices. The enhanced LLL image is obtained by accumulating time-domain correlated PSD images. Experimental results show that the visual effects, spatial frequencies and contrast of the enhanced image are significantly improved.