Dark2Light: multi-stage progressive learning model for low-light image enhancement.

Due to severe noise and extremely low illuminance, restoring from low-light images to normal-light images remains challenging. Unpredictable noise can tangle the weak signals, making it difficult for models to learn signals from low-light images, while simply restoring the illumination can lead to noise amplification. To address this dilemma, we propose a multi-stage model that can progressively restore normal-light images from low-light images, namely Dark2Light. Within each stage, We divide the low-light image enhancement (LLIE) into two main problems: (1) illumination enhancement and (2) noise removal. Firstly, we convert the image space from sRGB to linear RGB to ensure that illumination enhancement is approximately linear, and design a contextual transformer block to conduct illumination enhancement in a coarse-to-fine manner. Secondly, a U-Net shaped denoising block is adopted for noise removal. Lastly, we design a dual-supervised attention block to facilitate progressive restoration and feature transfer. Extensive experimental results demonstrate that the proposed Dark2Light outperforms the state-of-the-art LLIE methods both quantitatively and qualitatively.

[Measurement of path transverse wind velocity profile using light forward scattering scintillation correlation method].

A new method for path transverse wind velocity survey was introduced by analyzing time lagged covariance function of different separation sub-apertures of Hartmann wavefront sensor. A theoretical formula was logically deduced for the light propagation path transverse wind velocity profile. According to the difference of path weighting function for different sub apertures spacing, how to select reasonable path weighting functions was analyzed. Using a Hartmann wavefront sensor, the experiment for measuring path transverse velocity profile along 1 000 m horizontal propagating path was carried out for the first time to our knowledge. The experiment results were as follows. Path transverse averaged velocity from sensor had a good consistency with transverse velocity from the wind anemometer sited near the path receiving end. As the path was divided into two sections, the path transverse velocity of the first section had also a good consistency with that of the second one. Because of different specific underlaying surface of light path, the former was greater than the later over all experiment period. The averaged values were 1.273 and 0.952 m x s(-1) respectively. The path transverse velocity of second section and path transverse averaged velocity had the same trend of decrease and increase with time. The correlation coefficients reached 0.86.

[Stray light and bandpass correction in the spectral measurement for light emitting diodes].

The present paper introduces the stray light and bandpass correction methods for spectrometers. The line spread function of spectrometer is characterized by a He-Ne laser. Assuming that the spectrometer is a wavelength invariable system, the stray light distribution matrix is constructed by the derived line spread functions. The stray light correction matrix is then derived by matrix conversion from the stray light distribution matrix. The measured signals of the spectrometer are finally multiplied by the stray light correction matrix to correct the stray light errors. The bandpass functions of the spectrometer are characterized in three different wavelength ranges, respectively. And then three groups of bandpass correction coefficients are calculated accordingly. The calculation is divided into several steps. Given the measurement results at the target wavelength position and the ones on the neighbor bandwidths, the bandpass correction results are obtained by weight averaging of them. The bandpass correction coefficients are used as the weights. The two correction methods are applied to a multi-channel fast spectrometer to measure LEDs of different color. The results show that the stray light and the bandpass errors can be corrected effectively. The chromaticity coordinates of the LEDs are corrected by (-0.003, 0.007) for the maximum. Furthermore, the method introduced in this paper can reduce the application cost, simplify the calculation under a reasonable precision, and make the application of the correction easier.

[The analysis of the sampling modulation transfer function and the influence on the Gaussian spectra].

As the parameters of the photoelectric detector have important effects on the performance of the dispersive spectrometers, it is necessary to detail the discrete sampling process of the photoelectric detector array. In the present paper, the sampling model was setup, and the effects on the sampling results caused by the spatial frequency of the cosine signal, the width of the sampling pixel, and the initial phase of the sampling pixel position to the crest of the input cosine signal were discussed thoroughly in the frequency domain. By introducing the integral function, a general expression of the sampling modulation transfer function was given, and the concept and expression of the average sampling modulation transfer function was proposed. Since that expression eliminates the effect of initial phase, it is much more convenient to the practical applications. For the typical Gaussian spectrum produced by the dispersive spectrometer, the Fourier transform result of that spectrum was multiplied by the average sampling modulation transfer function to produce a functional expression of the modulation transfer function of the whole system. The average aliasing error of the sampling process was expressed as a function of spatial frequencies; the relationship between the peak value of the average aliasing error and the width of the Gaussian spectrum was discussed; and the critical value of the spectrum width to restore this spectrum precisely was proposed. That critical value is significant for providing guidance to the design and fabrication of dispersive spectrometers.

[A CMOS vertically integrated device for monochromatic spectrum detection].

A vertically integrated sensor based on standard CMOS process, which can detect the wavelength of the monochromatic spectrum, was designed, manufactured and tested. It took the advantage of the two-layer structure of the device which can sense short wavelength and long wavelength illumination (among near UV, visible and near IR) simultaneously, then through these two different responses, the final device response showed monotonically increasing with the wavelength, and the wavelength can be known through this monotonicity. First we introduced the basic principle of the device, then the ideas in test device design were given. Finally, the authors measured the manufactured device, and the QEs of the device and the monotonic relationship between the device and the wavelength of the monochromatic spectrum were given.

[Comparison of two methods to solve AR model parameters of line shape optimized maximum entropy spectral estimation].

Line shape optimized maximum entropy spectral estimation (LOMEE) is to remove the influence resulting from line shape by Fourier self-deconvolution and to model interferogram data in order to substitute AR model parameters for the spectral estimator. The proper calculation of AR model parameters in LOMEE is very important to the quality of recovered spectrum. Modified covariance method (MCOV) and Burg method were used to solve AR model parameters under different noise level. The simulation results of recovered spectrum were acquired with the two methods. The influences of AR order and signal-to-noise ratio on the simulation were compared. It is shown that MCOV is more excellent than Burg in LOMEE.

[Multispectral color sensor based on vertically stacked structure].

A novel multispectral color sensor based on vertically stacked structure is introduced. The advantages of this sensor include anti color aliasing, high spatial resolution, elimination of color interpolation and low-pass filter. Its basic principle lies on the silicon material's different of penetration depth of electromagnetic wave with different wavelength, i.e., blue light with short wavelength is mainly absorbed on the surface, while red light with longer wavelength is mainly absorbed at deeper location. The current research and development are related to two pixel structures: the buried pn junctions structure made by standard silicon process, and stacked amorphous silicon and its alloys thin films made by PECVD. The former one adopts standard silicon technology while the latter one adopts amorphous silicon, which has better optical performance, thus greater flexibility in design. The authors focused on the theoretical and experimental analyses of the spectrum mechanisms and output performances with different pixel structures, and discussed-the way for further research.