This paper presents a new hyperspectral image (HSI) classification method which is capable of automatic feature learning while achieving high classification accuracy. The method contains two major modules: the spectral classification module and the spatial constraint module. Spectral classification module uses a deep network named stacked denoising autoencoders (SdA) to learn feature representation of the data. Through SdA, the data are projected nonlinearly from its original hyperspectral space to some higher dimensional space where more compact distribution is obtained. An interesting aspect of this method is that it does not need a feature design/extraction process guided by human prior. The suitable feature for the classification is learned by the deep network itself. Superpixel is utilized to generate the spatial constraints to refine the spectral classification results. By exploiting the spatial consistency of neighborhood pixels, the accuracy of classification is further improved by a big margin. Experiments on the public datasets reveal the superior performance of the proposed method.

Different from the photometric images, depth images resolve the distance ambiguity of the scene, while the properties, such as weak texture, high noise, and low resolution, may limit the representation ability of the well-developed descriptors, which are elaborately designed for the photometric images. In this paper, a novel depth descriptor, geodesic invariant feature (GIF), is presented for representing the parts of the articulate objects in depth images. GIF is a multilevel feature representation framework, which is proposed based on the nature of depth images. Low-level, geodesic gradient is introduced to obtain the invariance to the articulate motion, such as scale and rotation variation. Midlevel, superpixel clustering is applied to reduce depth image redundancy, resulting in faster processing speed and better robustness to noise. High-level, deep network is used to exploit the nonlinearity of the data, which further improves the classification accuracy. The proposed descriptor is capable of encoding the local structures in the depth data effectively and efficiently. Comparisons with the state-of-the-art methods reveal the superiority of the proposed method.

Photovoltaic solar panels are effective energy sources during periods of bright sunlight. Excess energy can be stored for later use at night or on cloudy days. The decision to use the stored energy now or later depends largely on being able to predict the weather on different timescales. Short term prediction of stored energy is challenging due to the non-trivial I-V characteristic of the solar cell. The erratic nature of the weather makes long term predictive energy management difficult. In this paper, we address these issues based on data collected from a solar panel, as well as its relationship to observations made of the weather. We observe that prediction, based on fuzzy decision trees, reduces the energy error by 22% compared to a constant prediction equal to the average on the studied period. Thus, exploiting the fuzzy classification provided by a fuzzy decision tree is a good improvement compared to the baseline.

Just as color can be shouted in primary tones or whispered in soft pastel hues, so stereo can be shoved in your face or raised ever so gently off the screen plane. This paper reviews the problems with "in your face stereo" and demonstrates that "just enough reality" is both gentle and effective in achieving stereoscopy's fundamental goal: resolving the front-back ambiguity inherent in 2D projections. Something unexpected and perhaps fortuitous falls out: soft stereo means small on-screen disparities, and when on-screen disparities are small, crosstalk is perceived as blur rather than as ghosting. Blur, particularly if it occurs only in the background and foreground (to which regions it can be relegated by "center-of-interest compensation"), is, in contrast to ghosting, quite unobjectionable. Gracefully accepting crosstalk may enable zoneless autostereoscopic displays.

We address human factors and technology issues for the design of stereoscopic display systems that are natural and comfortable to view. Our title "just enough reality" hints at the contrast between the popularly perceived requirements for strict "virtual reality" and the expert's pragmatic acceptance of "sufficient reality" to satisfy the human interface requirements of real-world applications. We first review how numerous perceptions and illusions of depth can be exploited to synergistically complement binocular stereopsis. Then we report the results of our experimental studies of stereoscopy with very small interocular separations and correspondingly small on-screen disparities, which we call "microstereopsis." We outline the implications of microstereopsis for the design of future stereoscopic camera and display systems, especially the possibility of achieving zone-less autostereoscopic displays. We describe a possible class of implementations based on a nonlambertian filter element, and a particular implementation that would use an electronically switched louver filter to realize it.

We present a novel concept, vector restoration, for motion compensated predictive coding of video. Conventional predictive coders generate predictions of the input signal and encode the residual signal. Here, the predicted images are operated upon directly with the goal of restoring the original images at the decoder. The predicted images are processed in a piecewise manner using restoration functions chosen from a codebook. Compression is achieved by transmitting only an index of the chosen restoration function for each image region. We describe the manner in which the restoration functions are generated and chosen. Vector restoration is shown to outperform residual vector quantization and transform-based coders. This superiority is due to vector restoration's ability to adaptively exploit the non-trivial correlation between the predicted and residual images.

Stereoscopic sequence compression typically involves the exploitation of the spatial redundancy between the left and right streams to achieve higher compressions than are possible with the independent compression of the two streams. In this paper the psychophysical property of the human visual system, that only one high resolution image in a stereo image pair is sufficient for satisfactory depth perception, has been used to further reduce the bit rates. Thus, one of the streams is independently coded along the lines of the MPEG standards, while the other stream is estimated at a lower resolution from this stream. A multiresolution framework has been adopted to facilitate such an estimation of motion and disparity vectors at different resolutions. Experimental results on typical sequences indicate that the additional stream can be compressed to about one-fifth of a highly compressed independently coded stream, without any significant loss in depth perception or perceived image quality.>

We describe a nd demonstrate an ambient lllumlnation method for range Imaging In parallel with and using the same sensor that is used for Intensity Imaging. Our method automates the well known range-from-focus method. It makes use of our reaiizatlon that range can be calculated directly from a focus error signal: It is not necessary to actually focus the Image, only to measure the focus; error. We show theoretically and demonstrate experlmentaIly that the focus error distance can b e measured by time domain fourler analysls of the signal from a sensor that Is mlcroscoplcally oscillated (dithered) longitudinally in the space domain. In an approximation regime that our experiments show to be physically realistic, the ratio of the first two significant fourier amplitudes is proportional to the ratio of the focus error to the dither amplitude. Since the illuminatlon, the reflectance, and most technical noise sources affect all fourler amplitudes in an equal multiplicative fashion, the fourler amplitude ratio is a robust measure of the focus error to dither amplitude ratio. The constants needed to calculate absolute range images are the focal length of the lens, the sensor's mean position, and the dither amlpiitude. Experiments using a 50 mm W1.5 camera lens, a pinhole object, and a single detector dithered by an audio speaker demonstrate real-time range resolution of approxlmately 0.5 mun at a range of 350mm. Modelling lndlcatrs that the method, implemented with a sensor array, Is also applicable to extended objects. Plans for extending the experiments to real scenes are discussed.

A computer model is described that combines concepts from the fields of acoustics, linear system theory, and digital signal processing to simulate an acoustic sensor navigation system using time-of-flight ranging. By separating the transmitter/receiver into separate components and assuming mirror-like reflectors, closed-form solutions for the reflections from corners, edges, and walls are determined as a function of transducer size, location, and orientation. A floor plan consisting of corners, walls, and edges is efficiently encoded to indicate which of these elements contribute to a particular pulse-echo response. Sonar maps produced by transducers having different resonant frequencies and transmitted pulse waveforms can then be simulated efficiently. Examples of simulated sonar maps of two floor plans illustrate the performance of the model. Actual sonar maps are presented to verify the simulation results.