Author(s)

Wu Qiongfei, Xia Dingchun, Xie Qubo

Corresponding Author

Wu Qiongfei

ABSTRACT

In this paper, infrared image target extraction is realized based on discrete stationary wavelet transform and fractal dimension. Firstly, denoising and detail enhancement algorithms are designed in stationary wavelet domain at different resolutions to preprocess infrared image. Then, on this basis, the region of interest containing the target of interest (human and vehicle) is extracted by fractal dimension. Finally, the human and vehicle targets in the infrared image are extracted by global threshold segmentation method. In the process of algorithm design, db1 wavelet is used as the mother wavelet to speed up the running speed of the algorithm. Using fractal dimension to extract region of interest can improve the accuracy of subsequent target extraction. Further work considers summarizing the fractal dimension characteristics of different targets in infrared images to realize the effective segmentation of infrared images in more complex scenes.

KEYWORDS

Forward looking, infrared target, matching algorithm, matrix double transformation