The limited depth of field (DOF) is a common problem in imaging. It is typically desired to obtain images with a large DOF, and many methods have been reported in this regard, for example, wavefront coding, optical toe cutting, and focus scanning. However, changes in the aperture of a lens degrade image quality and increase the exposure time, and movements of detectors reduce the sharpness of the obtained images. Extended DOF images can also be realized by wavelet, contour wave, and focus area fusions of images captured with different focal lengths. The focus change is typically realized by the position changes of lenses. However, mechanical movements may shorten the life cycle and increase the volume and weight of an imaging system. Bai et al. found that a liquid crystal (LC) lens preserves their properties during state switching, which has been used to acquire video frames during the switching process to obtain high-quality extended DOF images. In this process, images can be easily acquired, but it takes a long time to calculate the DOF. Each frame in a video has a corresponding DOF that overlaps with other DOFs of the same video. It is not necessary to use all images for fusion, providing us with an idea to improve image acquisition efficiency by extending the DOF. In this study, aiming at the disadvantage that the existing method of acquiring extended DOF images by LC lenses requires a long time, a method for improving the efficiency of extended DOF imaging using the focus tuning property of LC lenses is proposed, which can significantly reduce the image acquiring time of extended DOF imaging.

The imaging system consists of a polarizer, an LC lens, a lens module, and a CMOS sensor. The DOF of the system can be calculated using the focal length of the main lens, the focal length of the LC lens, the F-number of the lens, the allowable speckle radius, and the object distance. First, the structure of the LC lens used in our experiment is described. The preservation of the lens property of the LC lens when it is driven to switch between positive and negative states is used to acquire images for the calculation. The voltage of the LC lens was scanned and measured. The power of the Lens is a function of time, and then, the relationship between the focus position of the system, DOF, and calculation time can be obtained. Then, by considering the DOF of each image, the DOF corresponding to each selected frame does not overlap. The number of images used can be significantly reduced, and the image processing time can be shortened. In this study, the complex value wavelet extended DOF fusion method is used to adaptively map the input image stack from color scalar and variance to a single-channel grayscale image, ensuring that channels with more details contribute more to the grayscale image. Finally, the difference between the images obtained using the proposed method and existing methods is compared in terms of the average gradient (AG), and the efficiency improvement performance of our algorithm is analyzed.

The camera module used in this study is Ming Mei MD50-T3, and the acquisition resolution is 1920×1080. 77 images were acquired during the state switching of the LC lens. Fig.10 shows the images focused on each object. With the proposed method, the number of images required for expanding the DOF is 24. Fig.11 shows the expanded DOF images fused with all 77 images and 24 images. There are a few subjective differences between the two image types. AG is used to evaluate the fusion effect of expanded DOF images. As shown in Fig.12, the same region of interest was selected for the fusion of the two image types and the corresponding source images. Image sharpness is sacrificed in the expanded DOF images obtained by fusing all images, and the AG of S₁ degrades by approximately 2%-6.5% relative to the source image. Although the clarity of the fusion image is reduced, the degree of degradation is acceptable compared with its significant advantage of DOF, and there is no obvious ambiguity phenomenon compared with the source image. The AG difference between S₁ and S₂ is mostly between 1% and 2%, which can be ignored. S₁ and S₂ can be thought to be equivalent to the visual effect. Image S₂, acquired by the fusion of the 24 selected images, achieves the same effect as the fusion of the 77 images. In this experiment, the number of images used for fusion is reduced to 31.2% of the original method, and the calculation time is reduced to 34.38% of the original method. This significantly improves the efficiency of image fusion and reduces the computation time.

A method of extending the DOF of images is proposed based on an LC lens. The preservation of the lens property of the LC lens when it is driven to switch between positive and negative states is used to acquire images for computation. The number of images used is significantly reduced by considering the DOF of each image, which shortens the image calculation time. Compared with existing DOF fusion methods based on LC lenses, the number of images used in this experiment is reduced to 31.2%, the calculation time is reduced to 34.38%, and the fusion efficiency is significantly improved. According to the experimental results in both subjective and objective evaluations, the extended DOF images have high quality.