Yujing Wu’s research while affiliated with Nanjing University of Science and Technology and other places

In this work, a fast imaging method based on sub-dark-field illumination in white-light interferometry is presented. Applying this method, we successfully resolved the surface profile of test structures even for a smooth surface with a roughness of the order of a few angstroms. The true-colour sequence band is presented, which provides the relationship between the surface colours and the optical path difference. Several characteristics (e.g., basic colour, image contrast, thickness variations, and so on) for the white-light interference images are discussed in detail. Pair images with sub-dark-field illumination in the principal interference fringe were also observed and analysed.

A fusion algorithm for infrared and multi-type images based on contrast pyramid transform (CPT) combined with Otsu method and morphology is proposed in this paper. Firstly, two sharpened images are combined to the first fused image based on information entropy weighted scheme. Afterwards, two enhanced images and the first fused one are decomposed into a series of images with different dimensions and spatial frequencies. To the low-frequency layer, the Otsu method is applied to calculate the optimal segmentation threshold of the first fused image, which is subsequently used to determine the pixel values in top layer fused image. With respect to the high-frequency layers, the top-bottom hats morphological transform is employed to each layer before maximum selection criterion. Finally, the series of decomposed images are reconstructed and then superposed with the enhanced image processed by morphological gradient operation as a second fusion to get the final fusion image. Infrared and visible images fusion, infrared and low-light-level (LLL) images fusion, infrared intensity and infrared polarization images fusion, and multi-focus images fusion are discussed in this paper. Both experimental results and objective metrics demonstrate the effectiveness and superiority of the proposed algorithm over the conventional ones used to compare.

The real-time fusion of multi-source images is a problem which is widely concerned but is hard to handle with because of the effectiveness and complexity of registration and fusion algorithms. In this paper, a matching rule based on image's ROI and a fusion rule based on LP transform is discussed to each frame of the video. To the top level image, the rule based on similarity judgment through histogram calculation and weighted image fusion is used; to the other levels, the gray values in fused image depend on the comparison results of two gradient values in source images. The fused images are colored according to different gray levels at last. In order to verify the effectiveness of algorithms, both visual effects and quantitative indicators are referred. The comparison results demonstrate that the adopted algorithms can achieve a satisfying result in all experimental images and thus can be applied as a real-time fusion way in practice.