Adaptive enhancement for infrared image using shearlet frame

Abstract

An infrared imaging sensor is sensitive to the variation of imaging environment, which may affect the image quality and blur the edges in an infrared image. Therefore, it is necessary to enhance the infrared image. To improve the image contrast and adaptively enhance image structures, such as edges and details, this paper proposes a novel infrared image enhancement algorithm in the shearlet transform domain. To avoid over-enhancing strong edges and amplifying noise in plateau regions, we linearly enhance the details on the high frequency components based on their structure information, and improve the global image contrast by non-uniform illumination correction on the low frequency component. Then we convert the processed low and high components into the spatial domain to obtain the final enhanced image. Experimental results show that the proposed algorithm could enhance the infrared image details well and produce few noise regions, which is very helpful for target detection and recognition.

Infer Thermal Information from Visual Information: A Cross Imaging Modality Edge Learning (CIMEL) Framework

“…Considering the relatively low signal-to-noise ratio (SNR) characteristics of IR images, some researchers adopted wavelet related algorithms to achieve better noise-reducing performance and edges preserving effects [11,12], while others separated the detail/edge information from the original IR image for different downstream tasks [13][14][15][16][17]. Furthermore, top-hat transform [18], gradient domain [19,20], shearlet domain [21,22] and frequency domain [23] have also been investigated for the IR edge/detail enhancement purpose. Some other related works including an improved unsharp mask algorithm [24], gradient distribution via Cellular Automata [25], morphological operators [26], all-optical upconversion imaging techniques [27], the iterative contrast enhancement method [28], and the gravitational force and lateral inhibition network [29].…”

Combining Regional Energy and Intuitionistic Fuzzy Sets for Infrared and Visible Image Fusion

“…The fusion methods based on multi-scale geometric analysis (MGA) are widely used in IR and VIS image fusion [ 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 ]. MGA tools can represent the images at different scales and different directions, and these characteristics are helpful to extract more detailed information of the images.…”

Non-Uniformity Correction of Infrared Images Based on Improved CNN With Long-Short Connections

“…where y ij is the response of the (i, j) detector, x ij is the real infrared radiation received by the (i, j) detector, g ij and o ij denote the gain noise and offset noise, respectively. Compared with visible images, infrared images have lower contrast, so it is necessary to enhance infrared images [5], [6], [7]. Meanwhile, infrared images have less detail information than visible images, so the image quality will be seriously affected if the noise removals are not enough.…”