Image segmentation, which has become a research hotspot in the field of image processing and computer vision, refers to the process of dividing an image into meaningful and non-overlapping regions, and it is an essential step in natural scene understanding. Despite decades of effort and many achievements, there are still challenges in feature extraction and model design. In this paper, we review the advancement in image segmentation methods systematically. According to the segmentation principles and image data characteristics, three important stages of image segmentation are mainly reviewed, which are classic segmentation, collaborative segmentation, and semantic segmentation based on deep learning. We elaborate on the main algorithms and key techniques in each stage, compare, and summarize the advantages and defects of different segmentation models, and discuss their applicability. Finally, we analyze the main challenges and development trends of image segmentation techniques.

A method is proposed to calibrate the long-wave infrared ultra-wide-angle camera in this paper. In addition, a novel calibration chessboard is designed and an advanced chessboard corner positioning method is adopted to improve the calibration precision. The designed calibration chessboard can achieve high thermal infrared contrast and exhibits outstanding stability, which is made of a thermoelectric semiconductor refrigeration device. The proposed subpixel corner positioning method can accurately locate the corners on the calibration chessboard according to the characteristics of the infrared image and the checkerboard pattern. Both the principle of the proposed infrared chessboard and the subpixel corner positioning procedure were presented, and the calibration experiment showed that the mean reprojection error and the root mean square error were reduced to 0.32 pixel and 0.39 pixel, respectively. Comparison studies were also performed to verify the calibration effect of the proposed method, and the possibilities of camera calibration error of the proposed method were analyzed.

A simple method is proposed to calibrate the fisheye distortion center, which synthesizes the algorithms of the ellipse fitting and Gaussian surface fitting. Through analyzing the imaging projection of the equidistant fisheye lens, the calibration principle is introduced in detail. From the experimental results, we find that ellipse fitting error and the relative error of Gaussian surface fitting are less than 0.06 pixels and 1.5%, respectively, and the proposed method has a higher calibration precision and stability compared with the methods in the inferences. Besides, the proposed method does not rely on any particular distortion model and other fisheye parameters, and only the image information of the calibration board is utilized, which reduces the calibration complexity and is suitable for the center calibration of any equidistant fisheye camera.

Aiming at the space target and two dominant background sources including the sun and earth-atmosphere system, their radiation models are established, and the background influence on the space target detection for the space-based infrared imaging system are discussed in the long wave infrared (8 to 14 μm) range. From the numerical calculation and analysis, we find that the earth-atmosphere system is the strongest background interference source and will form a big detection blind zone, followed by the sun, whose irradiance is about 1.6% of that from the earth-atmosphere system. Compared with these two intense radiations, the space target radiation acting on the space-based infrared detector is so weak that it may be submerged once the target appears in the earth-atmosphere or solar backgrounds. Thus, in order to reduce the influence of the background interferences, necessary measures must be taken to control the space-based infrared detector and its host platform in real time, such as adjusting the flight attitude, changing the detection angle, transferring operating orbit, and so on.

A distortion correction method for a fisheye image is proposed based on the virtual small-field (SF) camera. The correction experiment is carried out, and a comparison is made between the proposed method and the conventional global correction method. From the experimental results, the corrected image by this method satisfies the law of perspective projection, and the image looks as if it was captured by an SF camera with the optical axis pointing at the corrected center. This method eliminates the phenomena of center compression, edge stretch, and field loss, and the image character is more obvious, which benefits the afterward target detection and information extraction.