Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Graduate Research Compressive Sensing Microscopy for Nanomaterial Analysis Liyang Lu, Yun Li, Yibo Xu, Jianbo Chen, and Kevin F. Kelly Department of Electrical and Computer Engineering, Rice University, Houston, TX 77005 Compressive Hyperspectral Microscopy Abstract Compressive sensing (CS) is a novel imaging technology based on the inherent redundancy within most natural scenes. As the microscopic images of most nanomaterials may be considered sparse under certain transformations, it can take five to twenty times fewer measurements to acquire an image, and at the same time, achieve much higher sensitivity ratio compared to its raster-scanning counterparts. By applying CS in the 3-dimensional hyperspectral data acquisition processes, we develop a hyperspectral microscope system for the analysis of surface plasmon scattering spectra of metal nanostructures. A CS-based sum frequency generation imaging microscope is also demonstrated with the gold pattern sample illuminated by visible beam and infrared beam simultaneously. The possibility of applying a structurally simple and more economical scanning-mask instead of the digital micromirror device is also explored. Hyperspectral imaging in optical microscopy is of great importance in the study of various micro/nano-scale physical and chemical phenomena. By investigating the response of a specimen to electromagnetic waves as a function of both spatial position and wavelength, rich amount of information can be achieved for the specimen, e.g. the plasmonic scattering spectrum of each metal nano-structure within the field of view, or the fluorescence spectrum of each fluorescent molecule in a biomedical Traditional ways to acquire a hyperspectral data cube sample. However, the use of hyperspectral Hagen N, Kester RT, Gao L, Tkaczyk TS; Opt. imaging is limited by its high-dimensional Eng. 2012 Jun 13, 51(11). data acquisition technique. The hyperspectral data cube can be recovered from much fewer measurements than the total number of pixels. At the same time, an enhancement in sensitivity is achieved, because the spectrometer receives the sum of the light from half of all the pixels in each measurement. Examples of Image Sparsifying Transformations We constructed a hyperspectral microscope based on the principle of compressive sensing using a fiber optic spectrometer as the point detector. It can be used to study the scattering spectra of various metal nano-structures as well as many other kinds of samples. Introduction to Compressive Imaging Ag nanowires Au nanobelt “Polyol Synthesis of Silver Nanowires: An Extensive Parametric Study,” Sahin Coskun, Burcu Aksoy, and Husnu Emrah Unalan Crystal Growth & Design 2011 11 (11), 4963-4969 “A Tunable Plasmon Resonance in Gold Nanobelts,” Lindsey J. E. Anderson, Courtney M. Payne, Yu-Rong Zhen, Peter Nordlander, and Jason H. Hafner Nano Letters 2011 11 (11), 5034-5037 Ag nanowire spectrum DMD CS Sum Frequency Generation Microscopy A new sum frequency generation imaging microscope using compressive sensing has been developed for surface studies. Pseudorandom patterns were applied to a light modulator and reflected the sum frequency (SF) signal generated from the sample into a photomultiplier tube detector. The image of the sample was reconstructed using sparsity preserving algorithms from the SF signal. The results demonstrate the CS technique achieved 16 times the pixel density beyond the resolution where the raster scan strategy lost its ability to image the sample due to the dilution of the SF signal below the detection limit of the detector. Reconstructed 32x32 SFG images. Signal from Au stripes was bright and signal from Si substrate was dark. This work is done in collaboration with Steven Baldelli and Xiaojun Cai at UH Mask-based Compressive Imaging Limits of the DMD Tilting mirrors cause diffraction issues when used with coherent light source; Mirror size is not suitable for mid-infrared imaging. With the appropriate choice of mask and scanning patterns, the convolution of the shifted image creates similar compressive encoding as the patterns on the DMD. This not only eliminates the grating-like artifacts, it also allows for customizing the modulator element size to appropriate spectral range. Micromirrors Au nanobelt spectrum Sample imaged by CCD camera Rice Single-Pixel Camera Sparse signals can be recovered from far fewer incoherent linear measurements than the full number of raster measurements using an optical modulator. To perform the incoherent measurement, a spatial light modulator such as a digital micromirror device (DMD) is used to modulate the intensities of the image pixels, and lenses are used to summing the modulated light onto a point detector. The DMD from TI is an array of 2D mirrors that can tilt plus and minus 12 degrees rapidly. The image can be reconstructed very closely or exactly via the L1 norm minimization or total variation (TV) minimization technique. 5 slices of the normalized 128X128X100 hyperspectral data cubes between 500 nm and 620 nm. Resolution: 128X128; Spectral resolution: 1.2 nm; Compression ratio: 3:1; Integration time 300 ms. Advantages of CS imaging scheme Using single-element detector Economical realization of complicated imaging system, e.g. infrared imaging, hyperspectral imaging Compression during measurement leading to less amount of data IEEE Signal Processing Magazine, 25(2), pp. 83 - 91, March 2008 The reconstructed spectra clearly show a blue shift of the surface plasmon resonance peak along the Au nanobelt, which corresponds to the changing cross-sectional aspect ratio of the nanobelt. We can build on the ideas of CS multi-scale video (CS-MUVI) by creating a dual-scale mask pattern that when scanning results in both coarse and fine resolutions. The coarse resolution provides the preview necessary for the optical flow calculation needed to perform the full-scale L1 compressive reconstruction. Results for 5%, 10%, 20% and 50% compression on a test image seen on the right. 4-D microscopic system can be realized by applying CS-MUVI imaging framework to the compressive hyperspectral microscope. With a temporal dimension added, hyperspectral videos of living cells can be captured. Original Image This work is done in collaboration with Lindsey Anderson and Jason Hafner at Rice Picture and schematic diagram of SFG-CS microscope setup 32x32 Preview L1 Reconstruction This work is done in collaboration with Wotao Yin (Rice/UCLA) An example of dualscale mask pattern