In this report an infrared zoom optical system is discussed. This system shifts a single moving element to switch between two focal lengths(30mm/60mm), one for wide angle and the other for close-up. However, a conventional optics-only method cannot provide good imaging quality over a large depth of focus at each focus offset. To improve the imaging performance, we investigate a lens-combined modulated wavefront coding technology for extending the depth of focus. Instead of placing a phase mask at the pupil position like traditional wavefront coding does, all the element surfaces in the system contribute to achieving modulation transfer function (MTF) consistency over a large range of depth of focus under dual field-of-view settings. As a result, the new structure extends the depth of focus 6.5 times than that of the original system. We also demonstrate recovered images employing hyper-Laplacian priors with noise and artifacts suppressed. It is concluded that the novel structure can not only extend the depth of focus but also reduce the complexity of infrared optical system.
A hyper-Laplacian can model the heavy-tailed distribution of gradients in natural scenes well, which have proven effective priors for deconvolution and denoising. However, because of missing point spread function (PSF) information in the two-dimensional spatial domain of optical sparse aperture (OSA) systems, a hyper-Laplacian prior of single exposure cannot recover the missing information of images. The main focus of this paper is on combining hyper-Laplacian priors with a pupil and its rotated pupils to compensate PSF information and improve the image quality in OSA systems. A scheme of rotating the pupil that has double apertures is analyzed. The cost function relative to multiple degraded images and PSFs obtained by rotating the pupil is established. The alternating minimization algorithm consisting of two phases is implemented to acquire restored images. In one phase, the non-convex part of the problem is solved. In the other phase, the fast Fourier transforms (FFTs) are used to solve a quadratic equation in the frequency domain. Using the peak signal-to noise ratio (PSNR), a quantitative analysis is provided. Simulation results show that hyper-Laplacian priors combined with rotating pupils can restore images better than a hyper-Laplacian prior of single exposure in an OSA system. Taking spoke-square image as the test image, the PSNR is 28.34 dB with two rotations and 23.52 dB without rotation. Moreover, the numbers of rotating the pupil that lead to different changes of the image quality are demonstrated.
Visual object tracking plays a significant role in our daily life such as intelligent transportation and surveillance. However, an accurate and robust object tracker is hard to be obtained as target objects often go through huge appearance changes caused by deformation, abrupt motion, background clutter and occlusion. In this paper, we combine features extracted from deep convolutional neural networks pretrained on object recognition datasets with color name features and histogram of oriented gradient features skillfully to improve tracking accuracy and robustness. The outputs of the convolutional layers encode the senior semantic information of targets and such representations are robust to great appearance variations while their spatial resolution is too coarse to precisely locate targets. In contrast, color name features connected at the back of HOG features could provide more precise localization but are less invariant to appearance changes. We first infer the response of the convolutional features and HOG-CN features respectively, then make a linear combination of them. The maximum value of the result could represent the accurate localization of the target. We not only compare the tracking results of adopting a single feature alone, showing that the performance of them is inferior to ours, but also analyze the effect of exploiting features extracted from different convolutional layers on the tracking performance. What’s more, we introduce the adaptive target response map in our tracking algorithm to keep the target from drifting as much as possible. Extensive experimental results on a large scale benchmark dataset illustrates outstanding performance of the proposed algorithm.