There is a worldwide effort to apply 21st century intelligence to evolving our transportation networks. The goals of smart transportation networks are quite noble and manifold, including safety, efficiency, law enforcement, energy conservation, and emission reduction. Computer vision is playing a key role in this transportation evolution. Video imaging scientists are providing intelligent sensing and processing technologies for a wide variety of applications and services. There are many interesting technical challenges including imaging under a variety of environmental and illumination conditions, data overload, recognition and tracking of objects at high speed, distributed network sensing and processing, energy sources, as well as legal concerns. This paper presents a survey of computer vision techniques related to three key problems in the transportation domain: safety, efficiency, and security and law enforcement. A broad review of the literature is complemented by detailed treatment of a few selected algorithms and systems that the authors believe represent the state-of-the-art.
Lately, image personalization is becoming an interesting topic. Images with variable elements such as text usually
appear much more appealing to the recipients. In this paper, we describe a method to pre-analyze the image
and automatically suggest to the user the most suitable regions within an image for text-based personalization.
The method is based on input gathered from experiments conducted with professional designers. It has been
observed that regions that are spatially smooth and regions with existing text (e.g. signage, banners, etc.) are the
best candidates for personalization. This gives rise to two sets of corresponding algorithms: one for identifying
smooth areas, and one for locating text regions. Furthermore, based on the smooth and text regions found in
the image, we derive an overall metric to rate the image in terms of its suitability for personalization (SFP).
Automatic license plate recognition (ALPR) is an important capability for traffic surveillance applications, including toll
monitoring and detection of different types of traffic violations. ALPR is a multi-stage process comprising plate
localization, character segmentation, optical character recognition (OCR), and identification of originating jurisdiction
(i.e. state or province). Training of an ALPR system for a new jurisdiction typically involves gathering vast amounts of
license plate images and associated ground truth data, followed by iterative tuning and optimization of the ALPR
algorithms. The substantial time and effort required to train and optimize the ALPR system can result in excessive
operational cost and overhead. In this paper we propose a framework to create an artificial set of license plate images for
accelerated training and optimization of ALPR algorithms. The framework comprises two steps: the synthesis of license
plate images according to the design and layout for a jurisdiction of interest; and the modeling of imaging
transformations and distortions typically encountered in the image capture process. Distortion parameters are estimated
by measurements of real plate images. The simulation methodology is successfully demonstrated for training of OCR.
Image-based customization that incorporates personalized text strings into photorealistic images in a natural
and appealing way has been of great interest lately. We describe a semi-automatic approach for replacing text
on cylindrical surfaces in images of natural scenes or objects. The user is requested to select a boundary for the
existing text and align a pair of edges for the sides of the cylinder. The algorithm erases the existing text, and
instantiates a 3-D cylinder forward projection model to render the new text. The parameters of the forward
projection model are estimated by optimizing a carefully designed cost function. Experimental results show that
the text-replaced images look natural and appealing.
Natural language color (NLC) was initially developed as a web-based application and then deployed in one
Xerox print driver. NLC changes the image-editing paradigm from the use of curves, sliders, and knobs, to the
use of verbal text-based commands such as "make light green much less yellowish". The technology appeals
to a common user who has no expert knowledge in color science, and this naturally leads one to think about
its use in mobile devices. A prototype GUI design for a language-based color editing on iPhone platform will
be presented that uses several of its haptic interfaces (e.g. "slot-machine", shaking, swiping, etc.). A textual
interface is provided to select a color to be modified within the image and a direction of change for the
modification. A swipe interface is provided to select a magnitude and polarity for the modification. Actions on
the textual and swipe interface are converted to natural language commands that are in turn used to derive a
color transformation that is applied to relevant portions of the image to yield a modified image. The
modifications are displayed in real time to the user.
The availability of web and on-line image sharing services makes image personalization and customization a more
interesting topic. Nonetheless, designing a personalized image is a time-consuming task, requiring hours of work
by expert designers. Observing the potential opportunity to make the design process easier and more amenable
to ordinary users, we presented a semi-automatic tool for designing personalized images in the Electronic Imaging
(EI) symposium last year.1, 2
As a follow-up, we present several improvements to the original semi-automatic tool, for both text insertion
and text replacement on planar surfaces. We also describe our effort in implementing the tool as a true web-based
service, which eliminates the need for installation of any software or packages by the user. We believe that we
have made the technology of image personalization more friendly and accessible to ordinary users.
Image personalization is a widely used technique in personalized marketing,1 in which a vendor attempts to
promote new products or retain customers by sending marketing collateral that is tailored to the customers'
demographics, needs, and interests. With current solutions of which we are aware such as XMPie,2 DirectSmile,3
and AlphaPicture,4 in order to produce this tailored marketing collateral, image templates need to be created
manually by graphic designers, involving complex grid manipulation and detailed geometric adjustments. As
a matter of fact, the image template design is highly manual, skill-demanding and costly, and essentially the
bottleneck for image personalization.
We present a semi-automatic image personalization tool for designing image templates. Two scenarios are
considered: text insertion and text replacement, with the text replacement option not offered in current solutions.
The graphical user interface (GUI) of the tool is described in detail. Unlike current solutions, the tool renders
the text in 3-D, which allows easy adjustment of the text. In particular, the tool has been implemented in Java,
which introduces flexible deployment and eliminates the need for any special software or know-how on the part
of the end user.
Digital printing brings about a host of benefits, one of which is the ability to create short runs of variable,
customized content. One form of customization that is receiving much attention lately is in photofinishing
applications, whereby personalized calendars, greeting cards, and photo books are created by inserting text strings
into images. It is particularly interesting to estimate the underlying geometry of the surface and incorporate the
text into the image content in an intelligent and natural way. Current solutions either allow fixed text insertion
schemes into preprocessed images, or provide manual text insertion tools that are time consuming and aimed
only at the high-end graphic designer. It would thus be desirable to provide some level of automation in the
image personalization process.
We propose a semi-automatic image personalization workflow which includes two scenarios: text insertion
and text replacement. In both scenarios, the underlying surfaces are assumed to be planar. A 3-D pinhole
camera model is used for rendering text, whose parameters are estimated by analyzing existing structures in
the image. Techniques in image processing and computer vison such as the Hough transform, the bilateral
filter, and connected component analysis are combined, along with necessary user inputs. In particular, the
semi-automatic workflow is implemented as an image personalization tool, which is presented in our companion
paper.1 Experimental results including personalized images for both scenarios are shown, which demonstrate
the effectiveness of our algorithms.
Several color-imaging algorithms such as color gamut mapping to a target device and resizing of color images have traditionally involved pixel-wise operations. That is, each color value is processed independent of its neighbors in the image. In recent years, applications such as spatial gamut mapping have
demonstrated the virtues of incorporating spatial context into color processing tasks. In this paper, we
investigate the use of locally based measures of image complexity such as the entropy to enhance the
performance of two color imaging algorithms viz. spatial gamut mapping and content-aware resizing of
color images. When applied to spatial gamut mapping (SGM), the use of these spatially based local
complexity measures helps adaptively determine gamut mapping parameters as a function of image content
- hence eliminating certain artifacts commonly encountered in SGM algorithms. Likewise, developing
measures of complexity of color-content in a pixel neighborhood can help significantly enhance
performance of content-aware resizing algorithms for color images. While the paper successfully employs
intuitively based measures of image complexity, it also aims to bring to light potentially greater rewards
that may be reaped should more formal measures of local complexity of color content be developed.
Four color printing is the common way to render a color image to paper. This four color printing has certain
side effects, among them the metameric representation of colors. This means that a single visual color can be
generated through multiple different four color combinations. This is normally considered a problem, however,
the problem description can be inverted and information can be embedded in a printed color image that is
perceptually invisible under normal illumination, but revealed to an infrared imaging system. This means that
certain security aspects can now produced in an essentially
Substrates found in standard digital color printing applications frequently contain optical brightening agents
(OBAs). These agents fluoresce under near UV light and are predominantly intended to increase the perceived
paper white and thus create a paper look and feel which is preferred by customers. The fluorescence
phenomenon poses a considerable challenge in standard color management applications, however, the
problem description can be inverted and information can be embedded in a printed color image that is perceptually
invisible under normal illumination, but revealed via substrate fluorescence under UV illumination.
From a practical standpoint, the approach works with standard high brightness office-type papers and does
not require any special materials or media, or any modifications to the imaging path inside the machine. This
means that certain security aspects can now produced in an essentially cost-neutral way.
Color printer calibration is always performed in the presence of noise. A major part of this noise is due to the spatial
non-uniformity of the printer. In many cases, the spatial non-uniformity does not repeat the same pattern page by page.
Some calibration techniques use randomizing the locations of patches and averaging a large numbers of measurements to
reduce the noise. Since the non-uniformity is a kind of systematic errors, it should be estimated and corrected in a more
effective way. This presentation describes a method to provide a more accurate color calibration/characterization result
by estimating the spatial non-uniformity presented on the printed target page and applying a correction to the very same
page itself. Instead of defining the noise in the color measurement space, e.g. the CIE Lab space, the new method
specifies the noise in the printer-dependent color space, e.g. the CMY space. After the first-round color calibration, an
inverse transform, from CIE Lab to CMY, is derived from measurements of the entire printed target. The noise dC, dM,
or dY is determined as the difference between the original CMY values and the output of the inverse transform with the
measured Lab values as the transform input. Since color patch locations are stochastically arranged, any "noticeable"
spatial pattern of the noise is most likely due to the printer non-uniformity in the corresponding channel. The nonuniformity
can be estimated by spatially smoothing the noise terms and the result can be subtracted from the original CMY input for a second-round calibration to achieve higher color accuracy.
Color printer calibration is the process of deriving correction
functions for device signals (e.g., CMYK), so that the device can
be maintained with a fixed known characteristic color response.
Since the colorimetric response of the printer can be a strong function
of the halftone, the calibration process must be repeated for
every halftone supported by the printer. The effort involved in the
calibration process thus increases linearly with the number of halftoning
methods. In the past few years, it has become common for
high-end digital color printers to be equipped with a large number of
halftones, thus making the calibration process onerous. We propose
a halftone independent method for correcting color (CMY or CMYK)
printer drift. Our corrections are derived by measuring a small number
of halftone independent fundamental binary patterns based on
the 22 binary printer model by Wang et al. Hence, the required
measurements do not increase as more halftoning methods are
added. First, we derive a halftone correction factor (HCF) that exploits
the knowledge of the relationship between the true printer
response and the 22-model predicted response for a given halftoning
scheme. Therefore, the true color drift can be accurately predicted
from halftone-independent measurements and corrected correspondingly.
Further, we develop extensions of our proposed color
correction framework to the case when the measurements of our
fundamental binary patches are acquired by a common desktop
scanner. Finally, we exploit the application of the HCF to correct
color drift across different media (papers) and for halftoneindependent
spatial nonuniformity correction.
We propose a novel scanner characterization approach for applications requiring color measurement of hardcopy output in calibration, characterization, and diagnostics applications. The method is advantageous for common practical color printing systems that use more than the minimum of three colorants necessary for subtractive color reproduction; printing with cyan (C), magenta (M), yellow (Y), and black (K) is the most prevalent example we use in our description. The proposed method exploits the fact that for the scenarios in consideration, in addition to the scanner RGB values for a scanned patch, the CMYK control values used to print the patch are also available and can be exploited in characterization. An indexed family of 3D scanner characterizations is created, each characterization providing a mapping from scanner RGB to CIELAB for a fixed value of K, the latter constituting the index for the characterization. Combined together, the family of 3D characterizations provides a single 4D characterization that maps scanner RGB obtained from scanning a patch and the K control value used for printing the patch to a colorimetric CIELAB measurement for the patch. A significant improvement in the robustness of the method to variations in printing is obtained by modifying the K index to utilize the scanned output for a black-only patch printed with the corresponding K value instead of directly utilizing the control K value used at the printer. Results show that the proposed 4D scanner characterization technique can significantly outperform standard 3D approaches in the target applications.
Color printer calibration is the process of deriving correction functions for device CMYK signals, so that
the device can be maintained with a fixed known characteristic color response. Since the colorimetric
response of the printer can be a strong function of the halftone, the calibration process must be repeated for
every halftone supported by the printer. The effort involved in the calibration process thus increases
linearly with the number of halftoning methods. In the past few years, it has become common for high-end
digital color printers to be equipped with a large number of halftones thus making the calibration process
onerous . We propose a halftone independent method for correcting color (CMY/CMYK) printer drift. Our
corrections are derived by measuring a small number of halftone independent fundamental binary patterns
based on the 2×2 binary printer model by Wang et. al. Hence, the required measurements do not increase
as more halftoning methods are added. The key novelty in our work is in identifying an invariant halftone
correction factor (HCF) that exploits the knowledge of the relationship between the true printer response
and the 2×2 predicted response for a given halftoning scheme. We evaluate our scheme both quantitatively
and qualitatively against the printer color correction transform derived with the printer in its "default
state". Results indicate that the proposed method is very successful in calibrating a printer across a wide
variety of halftones.
We propose a novel scanner characterization approach for applications requiring color measurement of hardcopy output in printer calibration, characterization, and diagnostic applications. It is assumed that a typical printed medium comprises the three basic colorants C, M, Y. The proposed method is particularly advantageous when additional colorants are used in the print (e.g. black (K)). A family of scanner characterization targets is constructed, each varying in C, M, Y and at a fixed level of K. A corresponding family of 3-D scanner characterizations is derived, one for each level of K. Each characterization maps scanner RGB to a colorimetric representation such as CIELAB, using standard characterization techniques. These are then combined into a single 4-D characterization mapping RGBK to CIELAB. A refinement of the technique improves performance significantly by using a function of the scanned values for K (e.g. the scanner's green channel response to printed K) instead of the digital K value directly. This makes this new approach more robust with respect to variations in printed K over time. Secondly it enables, with a single scanner characterization, accurate color measurement of prints from different printers within the same family. Results show that the 4-D characterization technique can significantly outperform standard 3-D approaches especially in cases where the image being scanned is a
patch target made up of unconstrained CMYK combinations. Thus the algorithm finds particular use in printer characterization and diagnostic applications. The method readily generalizes to printed media containing other (e.g "hi-fi") colorants, and also to other image capture devices such as digital cameras.
The use of four process inks (CMYK) is common practice in the graphic arts, and provides the foundation for many output device technologies. In commercial applications the number of inks are sometimes extended beyond the process inks depending on the customers’ requirements, and cost constraints. In inkjet printing extra inks have been used to both extend the color gamut, and/or improve the image quality in the highlight regions by using "light" inks. The addition of "light" inks are sometimes treated as an extension of the existing Cyan or Magenta inks, with the Cyan tone scale smoothly transitioning from the light to the dark ink as the required density increases, or are sometimes treated independently.
If one is to treat the light ink as an extension of the dark ink, a simple blend can work well where the light and dark inks fall at the same hue angle, but will exhibit problems if the light and dark inks hues deviate significantly. The method documented in this paper addresses the problem where the hues of the light and dark inks are significantly different. An ink interaction model is built for the light and dark inks, then a composite primary is constructed that smoothly transitions from the light ink to dark ink, preventing the blended ink from over inking, while ensuring a smooth transition in lightness, chroma, and hue.
The method was developed and tested on an XES (Xerox Engineering Systems) ColorGraphx X2 printer, on multiple substrates, and found to provide superior results to the alternative linear blending techniques.
This paper presents a short historical perspective of color reproduction technology, where we are at present, and what challenges lie ahead. Today prevalent color management systems are based on pixel-wise processing of color data represented by 3 channels. We believe this does not adequately reproduce the color experience for the user, and that additional dimensions need to be considered. Three such dimensions are discussed: spectral, spatial, and goniometric. We believe the spectral representation plays an important role in color measurement and modeling; however a spectral image representation is likely not to be a mainstream technology due to the high cost incurred, and modest benefits that are appreciated only in certain niche applications. The incorporation of spatial context is likely to be of significant immediate benefit in commercial color imaging applications. Finally, the use of goniometric information in color reproduction is in its infancy, and bears promise. A brief overview of research efforts in each area is presented.
Monochrome devices that receive color imagery must perform a conversion from color to grayscale. The most common approach is to calculate the luminance signal from the three color signals. The problem with this approach is that the distinction between two colors of similar luminance (but different hue) is lost. This can be a significant problem when rendering colors within graphical objects such as pie charts and bar charts, which are often chosen for maximum discriminability.
This paper proposes a method of converting color business graphics to grayscale in a manner that preserves discriminability. Colors are first sorted according to their original lightness values. They are then spaced equally in gray, or spaced according to their 3-D color difference from colors adjacent to them along the lightness dimension. This is most useful when maximum differentiability is desired in images containing a small number of colors, such as pie charts and bar graphs. Subjective experiments indicate that the proposed algorithms outperform standard color-to-grayscale conversions.
Color device calibration is traditionally performed using one-dimensional per-channel tone-response corrections (TRCs). While one-dimensional TRCs are attractive in view of their low implementation complexity and efficient real-time processing of color images, their use severely restricts the degree of control that can be exercised along various device axes. A typical example is that 1-D TRCs in a printer can be used to either ensure gray balance along the C = M = Y axis or to provide a linear response in ΔE units along each of the individual (C, M and Y) axis but not both. This paper proposes a novel two-dimensional calibration architecture for color device calibration that enables significantly more control over the device color gamut with a modest increase in implementation cost. Results show significant improvement in calibration accuracy and stability when compared to traditional 1-D calibration.
Color device characterization involves deriving a mathematical description of the device response to a known input. This is known as the forward characterization transform. In the final application, this transform must be inverted to generate a mapping that determines the device input required for a desired response. This paper focuses on the inverse characterization transform for hardcopy devices. This can be discussed for two cases:
(1) Devices employing 3 channels
A colorimetrically unique inverse mapping exists provided the input signal is within the achievable domain of the device. When the forward transform is described by an analytic model, the inverse can be obtained by search-based techniques. When the forward transform is obtained empirically, the inverse transform is estimated by 3-D fitting or interpolation methods.
(2) Devices employing > 3 channels.
The inverse mapping is not colorimetrically unique, and therefore ill-posed. Additional constraints must be incorporated to ensure uniqueness. As an example, the case of CMYK printer characterization will be discussed. Constraints via undercolor removal and gray component replacement will be presented. Other methods that explicitly constrain CMYK combinations based on criteria such moire minimization will also be described.
For both cases, the problem of out-of-domain mapping and noise considerations will be discussed.
This paper presents a method for automatically extracting information about the source of a CMYK file by analyzing the image data. Several features are analyzed. These include an estimation of the type of undercolor removal and gray component replacement used to generate the image, and measures of image saturation and luminance. Together, these features provide a reasonable indication of the device for which the image was intended. While it is difficult to provide the exact source of an arbitrary file, the intention is to identify, with some degree of confidence, a probable class of devices for which the image was prepared (e.g. offset vs. laser vs. inkjet, etc.) so that the burden on the user to make this determination is reduced. Experiments performed to classify CMYK images from xerographic vs offset sources show promising results.
Natural pictures differ from synthetic graphics in many aspects, both in terms of visual perception and image statistics. As a result, image processing algorithms often behave differently on these two types of images. Classifying images and processing them using the method that is best to the image type may yield optimal results. In this paper, we propose to use image smoothness features to determine whether a scanned image was originally a synthetic graphic, or a natural picture. Synthetic graphics are usually very smooth. On the contrary, natural pictures are often noisier and texture rich. A classifier can therefore be built based on the measurement of the texture energy.
In this paper, a color correction system is embedded into a multiresolution representation with the goal of reducing the complexity of 3D look-up table transformations. A framework is assumed wherein the image undergoes a multiresolution decomposition, e.g. discrete wavelet transform, for the purpose of image compression or other processing. After the image is reconstructed from its multiresolution representation, color correction is usually required for rendering to a specific device. The color correction process is divided into two phases: a complex multidimensional transform (Phase 1), and a series of essentially 1-D transforms (Phase 2). Phase 1 correction is then moved within the multiresolution reconstruction process in such a way that a small subset of the image samples undergoes the multidimensional correction. Phase 2 correction is then applied to all image samples after the image is reconstructed to its full resolution. The recently proposed spatial CIELAB model is used to evaluate the algorithm. The computational cost incurred by the color correction is considerably reduced, with little loss in image quality.
A technique is proposed for estimating the surface of the color gamut of an output device, in 3D colorimetric space. The method employs a modified convex hull algorithm. This approach is shown to be more general, and more accurate, than existing techniques. Simple numerical metrics are derived from this surface description: namely the gamut volume in 3D space; and the percentage of colors from the Pantone Matching System which fall within the gamut.
Printer characterization and color correction are often complex transformations, and are derived with numerous measurements or printer models. There are many sources of errors in these transforms, including inaccuracies in lookup table approximation, errors in the printer model, noise in the data, and spatial and temporal non-uniformities in the printer. A method is proposed to increase the accuracy of an existing printer transform with a relatively small number of refinement measurements. A weighted linear least-squares regression technique is used to improve the fit of the printer response to the refinement data. The hypothesis is that a locally linear transform can adequately capture the difference between the true printer transform and its approximation. In contrast to existing approaches that only refine the individual C, M, Y, K responses, the proposed method attempts to account for cross-colorant interactions by using mixed colors in the refinement set. Furthermore, the refinement data is not restricted to lying on a regular grid, and can be freely chosen based on any a priori knowledge about the printer. The approach is tested for two related transforms: the characterization transform which maps CMYK to L*a*b*; and its inverse, the color correction transform that maps L*a*b* to CMYK. Results show an improvement in transform accuracy with a relatively small number of measurements.
A printer model is described for dot-on-dot halftone screens. For a given input CMYK signal, the model predicts the resulting spectral reflectance of the printed patch. The model is derived in two steps. First, the C, M, Y, K dot growth functions are determined which relate the input digital value to the actual dot area coverages of the colorants. Next, the reflectance of a patch is predicted as a weighted combination of the reflectances of the four solid C, M, Y, K patches and their various overlays. This approach is analogous to the Neugebauer model, with the random mixing equations being replaced by dot-on-dot mixing equations. A Yule-Neilsen correction factor is incorporated to account for light scattering within the paper. The dot area functions and Yule-Neilsen parameter are chosen to optimize the fit to a set of training data. The model is also extended to a cellular framework, requiring additional measurements. The model is tested with a four color xerographic printer employing a line-on-line halftone screen. CIE L*a*b* errors are obtained between measurements and model predictions. The Yule-Neilsen factor significantly decreases the model error. Accuracy is also increased with the use of a cellular framework.
An efficient algorithm for color image quantization is proposed based on a new vector quantization technique that we call sequential scalar quantization. The scalar components of the 3-D color vector are individually quantized in a predetermined sequence. With this technique, the color palette is designed very efficiently, while pixel mapping is performed with no computation. To obtain an optimal allocation of quantization levels along each color coordinate, we appeal to the asymptotic quantization theory, where the number of quantization levels is assumed to be very large. We modify this theory to suit our application, where the number of quantization 1evels is typically small. To utilize the properties of the human visual system (HVS), the quantization is performed in a luminance-chrominance color space. A luminance-chrominance weighting is introduced to account for the greater sensitivity of the HVS to luminance than to chrominance errors. A spatial activity measure is also incorporated to reflect the increased sensitivity of the HVS to quantization errors in smooth image regions. The algorithm yields high-quality images and is significantly faster than existing quantization algorithms.
In this paper, we propose a new technique for halftoning color images. Our technique parallels recent work in model-based halftoning for both monochrome and color images; we incorporate a human visual model that accounts for the difference in the responses of the human viewer to luminance and chrominance information. Thus, the RGB color space must be transformed to a luminance/chrominance based color space. The color transformation we use is a linearization of the uniform color space L*a* b* which also decouples changes between the luminance and chrominance components. After deriving a tractable expression for total- squared perceived error, we then apply the method of Iterated Conditional Modes (ICM) to iteratively toggle halftone values and exploit several degrees of freedom in reducing the perceived error as predicted by the model.
We investigate an efficient color image quantization technique that is based upon an existing binary splitting algorithm. The algorithm sequentially splits the color space into polytopal regions and picks a palette color from each region. At each step, the region with the largest squared error is split along the direction of maximum color variation. The complexity of this algorithm is a function of the image size. We introduce a fast histogramming step so that the algorithm complexity will depend only on the number of distinct image colors, which is typically much smaller than the image size. To keep a full histogram at moderate memory cost, we use direct indexing to store two of the color coordinates while employing binary search to store the third coordinate. In addition, we apply a prequantization step to further reduce the number of initial image colors. In order to account for the high sensitivity of the human observer to quantization errors in smooth image regions, we introduce a spatial activity measure to weight the splitting criterion. High image quality is maintained with this technique, while the computation time is less than half of that of the original binary splitting algorithm.
We apply the vector quantization algorithm proposed by Equitz to the problem of efficiently selecting colors for a limited image palette. The algorithm performs the quantization by merging pairwise nearest neighbor (PNN) clusters. Computational efficiency is achieved by using k- dimensional trees to perform fast PNN searches. In order to reduce the number of initial image colors, we first pass the image through a variable-size cubical quantizer. The centroids of colors that fall in each cell are then used as sample vectors for the merging algorithm. Tremendous computational savings is achieved from this initial step with very little loss in visual quality. To account for the high sensitivity of the human visual system to quantization errors in smoothly varying regions of an image, we incorporate activity measures both at the initial quantization step and at the merging step so that quantization is fine in smooth regions and coarse in active regions. The resulting images are of high visual quality. The computation times are substantially smaller than that of the iterative Lloyd-Max algorithm and are comparable to a binary splitting algorithm recently proposed by Bouman and Orchard.
Video Surveillance and Transportation Imaging Applications 2015
10 February 2015 | San Francisco, California, United States
Video Surveillance and Transportation Imaging Applications 2014
3 February 2014 | San Francisco, California, United States
Video Surveillance and Transportation Imaging Applications
4 February 2013 | Burlingame, California, United States
Mobile Computational Photography
4 February 2013 | Burlingame, California, United States
SC1131: Computer Vision and Imaging in Transportation Applications
This course introduces the attendee to applications in the transportation industry that employ imaging, computer vision, and video processing technologies. The class begins with a survey of key topics in transportation falling under three broad categories: safety, efficiency, and security. Topics include driver assistance, traffic surveillance and law enforcement, video-based tolling, monitoring vehicles of interest, and incident detection. The second part of the course provides a more in-depth treatment of state-of-art approaches to selected problems such as vehicle license plate recognition, vehicle occupancy estimation, speed enforcement, driver attention monitoring, and sensing of road and environmental conditions. Where necessary, background material on relevant computer vision concepts will be covered, such as image segmentation, object detection, classification, recognition, and tracking, and 3D camera geometry.