A general non-parametric technique is proposed for the analysis of a complex multimodal feature space and to delineate arbitrarily shaped clusters in it. The basic computational module of the technique is an old pattern recognition procedure: the mean shift. For discrete data, we prove the convergence of a recursive mean shift procedure to the nearest stationary point of the underlying density function and, thus, its utility in detecting the modes of the density. The relation of the mean shift procedure to the Nadaraya-Watson estimator from kernel regression and the robust M-estimators; of location is also established. Algorithms for two low-level vision tasks discontinuity-preserving smoothing and image segmentation - are described as applications. In these algorithms, the only user-set parameter is the resolution of the analysis, and either gray-level or color images are accepted as input. Extensive experimental results illustrate their excellent performance.

Mean shift: a robust approach toward feature space analysis

Nearly all super-resolution algorithms are based on the fundamental constraints that the super-resolution image should generate low resolution input images when appropriately warped and down-sampled to model the image formation process. (These reconstruction constraints are normally combined with some form of smoothness prior to regularize their solution.) We derive a sequence of analytical results which show that the reconstruction constraints provide less and less useful information as the magnification factor increases. We also validate these results empirically and show that, for large enough magnification factors, any smoothness prior leads to overly smooth results with very little high-frequency content. Next, we propose a super-resolution algorithm that uses a different kind of constraint in addition to the reconstruction constraints. The algorithm attempts to recognize local features in the low-resolution images and then enhances their resolution in an appropriate manner. We call such a super-resolution algorithm a hallucination or reconstruction algorithm. We tried our hallucination algorithm on two different data sets, frontal images of faces and printed Roman text. We obtained significantly better results than existing reconstruction-based algorithms, both qualitatively and in terms of RMS pixel error.

/pdf/limits-on-super-resolution-and-how-to-break-them-401rdhxruv.pdf

Limits on super-resolution and how to break them

Imaging vector fields has applications in science, art, image processing and special effects. An effective new approach is to use linear and curvilinear filtering techniques to locally blur textures along a vector field. This approach builds on several previous texture generation and filtering techniques[8, 9, 11, 14, 15, 17, 23]. It is, however, unique because it is local, one-dimensional and independent of any predefined geometry or texture. The technique is general and capable of imaging arbitrary two- and three-dimensional vector fields. The local one-dimensional nature of the algorithm lends itself to highly parallel and efficient implementations. Furthermore, the curvilinear filter is capable of rendering detail on very intricate vector fields. Combining this technique with other rendering and image processing techniques — like periodic motion filtering — results in richly informative and striking images. The technique can also produce novel special effects.

/pdf/imaging-vector-fields-using-line-integral-convolution-rq9l8mskf1.pdf

Imaging vector fields using line integral convolution

We introduce a general technique for "colorizing" greyscale images by transferring color between a source, color image and a destination, greyscale image. Although the general problem of adding chromatic values to a greyscale image has no exact, objective solution, the current approach attempts to provide a method to help minimize the amount of human labor required for this task. Rather than choosing RGB colors from a palette to color individual components, we transfer the entire color "mood" of the source to the target image by matching luminance and texture information between the images. We choose to transfer only chromatic information and retain the original luminance values of the target image. Further, the procedure is enhanced by allowing the user to match areas of the two images with rectangular swatches. We show that this simple technique can be successfully applied to a variety of images and video, provided that texture and luminance are sufficiently distinct. The images generated demonstrate the potential and utility of our technique in a diverse set of application domains.

/pdf/transferring-color-to-greyscale-images-2q619t61ig.pdf

Transferring color to greyscale images

This paper presents a methodology for evaluation of low-level image analysis methods, using binarization (two-level thresholding) as an example. Binarization of scanned gray scale images is the first step in most document image analysis systems. Selection of an appropriate binarization method for an input image domain is a difficult problem. Typically, a human expert evaluates the binarized images according to his/her visual criteria. However, to conduct an objective evaluation, one needs to investigate how well the subsequent image analysis steps will perform on the binarized image. We call this approach goal-directed evaluation, and it can be used to evaluate other low-level image processing methods as well. Our evaluation of binarization methods is in the context of digit recognition, so we define the performance of the character recognition module as the objective measure. Eleven different locally adaptive binarization methods were evaluated, and Niblack's method gave the best performance.

Goal-directed evaluation of binarization methods

Digital image processing (2nd ed.)

The Programmer's Imaging Kernel System (PIKS) is an application program interface (API) for image processing. It is one of three parts of a standard for Image Processing and Interchange being developed by the International Standards Organization (ISO) and the International Electrotechnical Commission (IEC). This paper presents an overview of the API; companion papers discuss the imaging architecture and image interchange parts of the standard. PIKS contains a rich set of operators, tools, and utilities. PIKS operators are functional elements that perform manipulations of images or of data objects extracted from images in order to enhance, restore, or assist in the extraction of information from images. These operators range from primitive operators such as convolution and histogram generation to complex, higher level operators such as adaptive histogram equalization and texture feature extraction. PIKS tools are elements that create data objects to be used by PIKS operators, e.g., the generation of filter transfer functions. PIKS utilities are elements that perform basic mechanical implementation tasks such as extracting pixels from an image. PIKS provides a fundamental operator model that supports match point translation of images prior to processing, image-related region-of-interest processing control, image/operator coordinate index assignment, and the ability to define reusable chains of operators.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

William K. Pratt

Papers

Digital image processing (2nd ed.)

Overview of the ISO/IEC programmer's imaging kernel system application program interface