Infrared maritime target recognition plays a significant role in the field of maritime search and maritime rescue. However, complex maritime conditions and illumination interference decrease the quality of maritime infrared images. For example, the backlight maritime condition can result in maritime targets with a negative contrast and make them disappear in the maritime background. How to enhance the contrast of maritime infrared targets and improve the quality of dim targets in the backlight maritime condition is the significant basis of maritime target detection and recognition. Furthermore, studying effective methods for detecting maritime targets in the backlight maritime condition is a significant research direction, and conventional target detection algorithms often lead to a high false alarm rate and low detection rate. We hope that the proposed enhancement algorithm can outstand maritime targets in the backlight maritime condition and the proposed target detection algorithm can detect maritime targets in backlight maritime conditions with a high accuracy rate.

We have tested the histograms of multiple maritime infrared images in the backlight condition and studied the characteristics of these histograms. Then, we find that the histograms can reflect the characteristic of a local peak, which represents the large background in the backlight maritime condition. We limit the maximum proportion in the histograms and obtain a novel histogram. The novel histogram is equalized, and it can prevent illumination interference and improve image quality. In addition, we quantize and extract the edge information in a backlight maritime image, and the edge information is fused with a suitable proportional parameter in the middle result of novel histogram equalization, which makes the enhancement result have high contrast and more detailed information. Conventional methods for target detection such as the local contrast method (LCM) are used in target detection with positive contrast by a target detection unit with a single scale. In the proposed local contrast method with multiscale for target recognition in backlight condition (LCMMBC) algorithm, we establish the target detection unit with multiple scales and define the local contrast saliency between the local targets and the local background in a negative contrast condition, and some significant procedures such as moving steps, pooling strategy, and threshold selection are discussed. Finally, the pseudocode and the implementation process of the LCMMBC algorithm are described.

Infrared maritime images have the characteristics of large background with a low gray value and target region with a negative contrast which can easily disappear in the large background (Fig. 1). The histograms of the maritime infrared images in the backlight condition have a local peak (Fig. 2), and the local peak is the high proportion of pixels in image histograms, which represents the large background in backlight maritime infrared images. The diagram of the original histogram modification includes the maximum proportion's limit and normalization of the novel proportion of pixels (Fig. 4). The result of the novel histogram equalization (Fig. 5) shows that the illumination condition is reasonably adjusted compared with the original image (Fig. 1). However, the middle result misses some details to some degree, and we need to add some edge information. The edge information's structural elements, quantification model, and extraction method are described in this paper (Fig. 6), which can reflect the gray value variation around the central pixel. At last, the result of histogram equalization with plateau limit and edge fusion (HEPLEF) algorithm has high local contrast, and the illumination of the enhancement result is uniform. In particular, dim maritime targets are highlighted by this proposed algorithm (Fig. 9). From the objective image quality assessment, it can be seen that the average gradient of the enhancement result is increased by more than two times than that of the original image (Table 1), and the local contrast gain is increased by more than two times than that of the original image (Table 2). The image assessment standard reflects that the HEPLEF algorithm can enhance the details of infrared maritime images and improve the contrast of infrared maritime images effectively. The support of the maritime target enhancement for target detection is also studied, and a three-dimensional diagram of local contrast saliency is used in the comparison between the enhancement result and the original image. Furthermore, the enhancement result increases the suspected target region's local contrast obviously (Fig. 18). The performance of the proposed LCMMBC algorithm is also tested, and we hope that the proposed algorithm can obtain a higher detection rate and lower false alarm rate. The experimental result shows that the proposed algorithm achieves a detection rate of 99.8% and a false alarm rate of 23.4%, respectively (Table 3), which shows better performance than other algorithms.

In this study, two novel algorithms called HEPLEF and LCMMBC are used for infrared maritime image enhancement and infrared target detection in the backlight condition, respectively. The HEPLEF algorithm can be applied in infrared maritime images with a large maritime background and dim maritime targets, and the enhancement result reflects that the contrast of the entire image is improved, and the target details are highlighted. The HEPLEF algorithm has characteristics of fewer input parameters and chief calculation. LCMMBC algorithm is suitable for maritime targets with a negative local contrast, and the performance of the LCMMBC algorithm is robust. In the principle of the LCMMBC algorithm, the disadvantages are the size of the minimum detection unit relies on the factual target size in the infrared image, and the selection of the minimum target detection size and local contrast saliency threshold of the LCMMBC algorithm is sometimes difficult.