Adaptive NUC algorithm for uncooled IRFPA based on neural networks

With the application of stitching technology in large-pixel-array CMOS image sensors, the problem of non-synchronized output signals from pixel array bilateral driver circuits has become progressively more serious and has led to the DC perforation of bilateral driver circuits, while conventional clock tree synchronization design methodology does not apply to stitching technology. Therefore, this paper analyses reasons for the inconsistency in the output signals of bilateral driving circuits and proposes a synchronous driving method applicable to stitching pixel arrays based on the idea of on-chip output signal delay detection and calibration. This method detects and corrects the non-synchrony of the row driver output signals on both sides according to changes in the operating environment of the chip. This method is characterized by a simple structure and high reliability. Finally, based on the 55 nm stitching process, simulations are carried out in a CMOS image sensor with a chip area of 77 mm × 84 mm to verify that this method is feasible. This large image sensor with a 150 M pixel array has a frame rate of over 10 FPS.

In this paper, we introduce a novel nonuniformity correction (NUC) algorithm for infrared focal-plane array (IRFPA). It is based on layers technique. First, the Rolling Guidance Filter (RGF) is utilized to decompose the raw IR image into a low frequency part and a high frequency part. Then, an adaptive temporal high-pass filter is utilized to filter the high frequency part by making use of the gradient and amplitude of it to estimate the Fixed Pattern Noise (FPN). The proposed scheme use the frames with large displacement to estimate the FPN to alleviate the ghosting artifacts in case of scene moves slowly. At Last, the estimated FPN is subtracted from the pristine image to obtain the correct result. Experiments with synthetic and real IR video demonstrate that the proposed method has better NUC performance and less artifacts than the state-of-the-arts.

An adaptive correction algorithm is demonstrated based on weak measurement, which introduces a feedback and an additional interaction to the system and can dynamically adjust the operating point in accordance with the condition of the estimated phase change. Two schemes, fast adaptive correction and slow adaptive correction, are proposed for different conditions of the modulation device. Fast adaptive correction scheme can realize a real-time correction and maintain the high sensitivity. Slow adaptive correction scheme, as a supplement, can correct the distortion of the measured parameter by changing the measuring period. These two schemes are useful for high precision phase estimation with time in modern physics and practical applications, including, but not limited to, timing synchronization, accurate distance measurement, and gravity wave detection. Moreover, we discuss the deviation of the adaptive correction for considering system noise in practical measurement.

Low-contrast or uneven illumination in real-world images will cause a loss of details and increase the difficulty of pattern recognition. An automatic image illumination perception and adaptive correction algorithm, termed as GLAGC, is proposed in this paper. Based on Retinex theory, the illumination of an image is extracted through the discrete wavelet transform. Two features that characterize the image illuminance are creatively designed. The first feature is the spatial luminance distribution feature, which is applied to the adaptive gamma correction of local uneven lighting. The other feature is the global statistical luminance feature. Through a training set containing images with various illuminance conditions, the relationship between the image exposure level and the feature is estimated under the maximum entropy criterion. It is used to perform adaptive gamma correction on global low illumination. Moreover, smoothness preservation is performed in the high-frequency subband to preserve edge smoothness. To eliminate low-illumination noise after wavelet reconstruction, the adaptive stabilization factor is derived. Experimental results demonstrate the effectiveness of the proposed algorithm. By comparison, the proposed method yields comparable or better results than the state-of-art methods in terms of efficiency and quality.

The microwave radiometer MTVZA-GY on board the Russian polar orbiting meteorological satellite Meteor-M N2 is briefly described. Errors and biases in antenna brightness temperatures are documented. Observation errors are found to be dependent on the solar angles. An adaptive bias correction technique for MTVZA-GY antenna temperatures is motivated and developed. The technique accounts for the solar angles and sequentially assimilates observed minus simulated radiances in a perpetual 24 h cycle in order to estimate up-to-date correction coefficients defined to be functions of the zenith and azimuth solar angles. The simulated radiances are computed by the RTTOV radiative transfer model from three-dimensional numerical weather prediction fields. The correction technique is implemented for atmospheric temperature and humidity sounding channels of MTVZA-GY. The corrected observations are shown to be significantly more accurate as compared with raw antenna temperatures and with observations that undergo simpler and more traditional corrections. The accuracy of corrected MTVZA-GY observations is compared with the accuracy of AMSU-A and MHS data.

The spatial non-uniformity in the photo-response of the detectors in the array on the focal-plane array (FPA) Infrared imaging systems restricted the infrared applications. In this paper, we improve the method of temporal high-pass filter for the complex real scene sequence. Firstly, it adopts the one point non-uniformity correction algorithm which calibrates the FPA at distinct temperatures by use of flat-field data generated from a black-body radiation source. It realized simply and it compensated for the spatial non-uniformity coarsely. After this step, the convergent time of the temporal high-pass algorithm was reduced, and the grade of ghost-shading was alleviated. And then, we analyze the pixels of images and classify them to two categories. One is the changed pixels, another is the stillness pixels. For different kinds of pixels, deal with different strategies. For the changed pixels, estimate the offset with the temporal high-pass filter algorithm. For the stillness pixels, estimate the offset with the iterative steps. This strategy reduced the grade of scene-vanishing when scene was stillness, and the grade of ghost-shading when target moving fast after stillness. Testing on the real infrared image sequence, the experiments showed that this method was very promising.

One of the most important preprocessing techniques is image Contrast Enhancement which is a technique that improve visual quality of an input image becoming more suitable for human analysis and perception. Numerous researches have been already developed for enhancement of Medical image for their important application, Traditional Gamma Correction is found to be one of the simplest technique for the contrast enhancement of medical image. This technique uses a set of varying parameter (y, c) which adjust effectively gray value intensity of the input image. Otherwise, Adaptive Gamma Correction technique have been appeared and have been proved its effectiveness by the use of adaptive (y, c) parameter which are determined adaptively from statistical information of input image. Traditional Gamma Correction (TGC) and Adaptive Gamma Correction (AGC) have been applied on three brain MRI modalities for patient with Multiple Sclerosis pathology. Qualitative and quantitative results are presented to illustrate the comparison of TGC and AGC to enhance the contrast of MRI images for better appearance of normal tissue and diseased tissue affected by MS pathology.

Nowadays, convergence and ghosting artifacts are common problems in scene-based nonuniformity correction (NUC) algorithms. In this study, we introduce the idea of space frequency to the scene-based NUC. Then the convergence speed factor is presented, which can adaptively change the convergence speed by a change of the scene dynamic range. In fact, the convergence speed factor role is to decrease the statistical data standard deviation. The nonuniformity space relativity characteristic was summarized by plenty of experimental statistical data. The space relativity characteristic was used to correct the convergence speed factor, which can make it more stable. Finally, real and simulated infrared image sequences were applied to demonstrate the positive effect of our algorithm.

This research presents the performance evaluation of two retina-like preprocessors for imaging detector arrays. In particular, the preprocessors are an adaptive neural network gain and level controller and a temporal high pass filter. These preprocessors were developed by Scribner et. al. at the Naval Research Laboratory for the purpose of chip-level non uniformity correction associated with InfraRed Focal Plane Arrays. Three performance evaluations are described. The first evaluation involves the determination of a spatio-temporal steady state transfer function (very similar to Modulation Transfer Function). The second involves the processor reduction of fixed pattern noise. Finally, the processor is evaluated on its capacity to reduce temporal l/f noise. The performance of the preprocessors are compared and contrasted with a discussion of their advantages and disadvantages.

A robust neural network (NN) output feedback scheme is proposed for the motion control of rigid robots. A dynamic NN observer is presented to estimate the joint speeds. The stability of a closed-loop system composed of a robot, a NN observer, and a NN controller is proven. The NN weights in both the observer and the controller are tuned online, with no off-line learning phase required. Most importantly, we can guarantee the boundness of the estimated velocities, the position tracking errors, and the NN weights. Also no exact knowledge of the robot dynamics is required so that the NN controller is model-free and so applicable to any type of rigid robot. When compared with adaptive-type controllers, we do not require persistent excitation conditions, linearity in the unknown system parameters, or the tedious computation of a regression matrix. Thus the new NN approach represents an improvement over adaptive techniques.

In this paper, an encoder and decoder system is proposed using Bose-Chaudhuri-Hocquenghem (BCH) double-error-correcting and triple-error detecting (DEC-TED) with emerging memories of low power and high decoding efficiency. An adaptive error correction technique and an invalid transition inhibition technique is enforced to the decoder. This is to improve the decoding efficiency and reduce the power consumption and delay. The adaptive error correction gives high decoding efficiency and invalid transition technique reduce the power consumption issue in conventional BCH decoders. The DEC-TED BCH decoder combines these two techniques by using a specific Error Correcting Code Clock and Flip Flops. This technique provides an error correcting encoder and decoder solution for low power and high-performance application using emerging memories. The design simulated in Xilinx FPGA using ISE Design Suite 14.5.

Aberration variations severely degrade retinal imaging in small animal eyes. Previously, the approach of a goggle lens with a matching corneal index was proposed to overcome the on-axis resolution limit of static imaging systems, which allows the use of the full eye pupil. But this technique didn’t address the problem of the large power variation, and the ensuing aberration on and off-axis, when dealing with small animal eyes. In this study, we present the concept of a tunable goggle lens, designed to compensate individual ocular aberration for different rodent eye powers. Ray tracing evidences that lens-fitted goggles permit, not only to adjust individual eyes power, but also to surpass conventional adaptive correction technique over large viewing angle, provided a minimum use of two spaced liquids. We believe that the overlooked advantage of the 3D lens function is a seminal finding for further technological advancements in widefield retinal imaging.

Underwater applications require clear underwater images, however, blurring and contrast degradation in underwater images are caused by absorption, dispersion, and water mass. To address the limitations of existing color correction and deblurring algorithms, this study proposes a fusion-based image enhancement algorithm. Two images obtained by adaptive channel defogging technique and restricted histogram mapping interval correction technique are used as two inputs to the fusion framework to obtain a new underwater image. We tested our technique with a public dataset. The experimental evidence shows that the technique proposed in this paper can improve the quality and detail of blurred underwater images. This paper also demonstrates that the method has superior deblurring and color correction capabilities than other existing algorithms.

Subject of study. Developed methods for synthesis of freeform illumination optics (ellipsoid, Monge-Ampere, string, and flow line methods) do not contain a correct description of the transformation of input and output wavefronts determined by the wave aberration function, do not consider the image quality criteria, and are not suitable for the calculation of imaging systems. The most common methods for synthesis of freeform elements of imaging systems (simultaneous multiple surface and ray-mapping methods) are also not ubiquitous because they are aimed at realization of a specific system configuration and, at least, require modification to solve a specific task. The largest complication in the synthesis stages of systems of imaging optics is the determination of the initial system scheme. Design of a method for synthesis of optical systems that does not require a starting point is a highly relevant objective of applied optics. A method allowing the synthesis of freeform optical surfaces for imaging systems using an artificial neural network is presented in this paper. Method. A neural network was built upon radial-basis functions and uses the geometric mapping method to generate the coordinates of the desired optical surface during backward ray tracing. The neural network was trained on different numbers of optical systems, which were automatically generated with variations in structural parameters and positions of the object and image. Main results. The method was demonstrated to operate effectively using more than 17,000 training systems at the ray set dimension of M=4096. The model of a flat-hyperbolic lens calculated using this method has a wave aberration of W<5×10−2 for wavelengths within the diffraction tolerance. Practical significance. The developed method can be used to design high-performance imaging systems.

The article describes the application of artificial neural networks and residual classes in the tasks of hardware-software implementation of neural networks. The comparison of existing software realizations of artificial neural networks is made. It is shown that a more effective implementation is hardware-implemented neural networks on the basis of a programmable logic device (PLD) type FPGA of Xilinx company. To achieve greater efficiency of training and calculations it is proposed to use the system of residual classes. The article shows the results of modeling finite ring neural networks (FRNN) on the basis of FPGA with minimal hardware costs and acceptable performance. For practical approbation of the results, a model of the neural network of adaptive resonance was chosen, its adaptation for implementation on the basis of PLD type FPGA was carried out. The developed neural network is trained for the classification of input vectors of images, testing is performed, which showed 100% quality of classification of input data at their noise (up to 15%).