This PDF file contains the editorial “Editorial: Progress or Not?” for JEI Vol. 3 Issue 02

Power cables present a persistent threat to helicopter pilots and have been responsible for a number of crashes in recent years. Active sensors are currently being developed to detect obstacles such as exposed power wires. However, the ability of passive sensors, such as forward-looking infrared (FLIR) systems currently employed on many rotary-wing aircraft, to detect such wires remains unclear. This study details observations, assumptions, and models that support the conclusion that wires are essentially reflective in nature, thus their signatures on passive thermal sensors depend principally on their surroundings. Because of this irregular nature and inconsistent appearance, predicting the performance of FLIRs in detecting wires is difficult and the utility of FLIRs as wire warning sensors is open to question.

A new system for the recognition of objects that are both complex and colored is presented. The transformation of the color cluster is designed to obtain a color set that makes it possible to use the color cluster information for n-tuple classification. Generally, if two objects have different color patterns, their color clusters will be different. Therefore, the color cluster is an important characteristic for colored object recognition. Further processing takes place, where some color planes are sliced from the color set, and every plane is sampled separately and applied to a classifier based on the n-tuple technique ofpattern recognition. The color cluster called pure color information is position invariant. Therefore, the system has one great advantage of position-invariant inspection for colored objects. The system interrogation time is always the same for whatever the complexity of the object, which is another advantage of the technique for the inspection of complex objects. Preliminary results using these techniques have shown promise in the classification and inspection of various artifacts.

A technique for the fast [O(N)] decomposition of images with respect to Zernicke polynomials is presented. The technique is used to compare images that are equivalent up to a rotation, to correct images misaligned by a rotation, to place symmetric images into a known orientation, and for pattern recognition applications in which rotation-invariant image features are desired. Zernicke polynomials form a class of rotation-invariant polynomials defined on the unit disk |z| ≤ 1. Although the generation of the coefficients of the expansion of an image in terms of these polynomials seems computationally inefficient, we show how the O(N) wavelet transform can be used to replace the Zernicke-image inner products by a component-wise product of the wavelet transforms of the Zernicke polynomials and the wavelet transform of the image. We show how the Zernicke decomposition of an image leads to the correction of rotational misalignment of magnetic resonance images, and how a wide class of centerline-symmetric images (such as magnetic resonance images of the head) can be brought into a uniform alignment using the Zernicke technique.

A gray-tone image taken of a real scene will contain inherent ambiguities due to light dispersion on the physical surfaces. The neighboring pixels may have very different intensity values and yet they may represent the same surface region. A fuzzy set theoretic approach for representing, processing, and quantitatively evaluating the information in gray-tone images is presented in this paper. The gray-tone digital image is mapped into a two-dimensional array of singletons called a fuzzy image. The value of each fuzzy singleton represents the degree to which a pixel intensity can be associated with some vaguely defined visual property γ. For illustrative purposes, the visual properties related to the notion of a uniform surface are investigated. The inherent ambiguity in the surface information can be modified by performing a variety of fuzzy mathematical operations on the singletons. Once the fuzzy image processing operations are completed, the modified fuzzy image can be converted back to a gray-tone image representation. The ambiguity associated with the processed fuzzy image is quantitatively evaluated by measuring the uncertainty present both before and after processing. Computer simulations are presented to illustrate some of these notions.

A neural network approach for image restoration is presented. The proposed method is based on a neural network with hierarchical cluster architecture (NNHCA), one of the recently emerged neural networks with sophisticated architectures. The method is motivated by the universally accepted concept in digital image processing that natural image formation is a local process. Therefore, the inverse problem of image restoration can be expressed by a globally coordinated local parallel processing (GCLPP) model. The GCLPP model can be readily realized by NNHCA. By utilizing the symmetric positive-definite quadratic structure of the formulation, a model-based local neuron evaluation algorithm is proposed. The algorithm significantly increases the convergence speed of restoration compared with previously proposed neural computing methods. A coordination scheme is also introduced to systematically resolve conflicting boundary conditions in the problem formulation. Visual examples are given to demonstrate that the proposed method not only produces good restoration results, but also provides a genuine parallel processing structure that ensures computationally feasible space domain image restoration.

Recently, many techniques have been introduced to improve on the reconstruction and enhancement of biomedical images, in particular, positron emission tomography. Often these techniques either perform poorly in the presence of noise or are computationally expensive. Nonintensive computational techniques are presented that aim to reduce noise while preserving edges working in conjunction with the filtered convolution back-projection method of reconstruction. First we present an adaptive windowing technique that bases window size on the count value. Then a maximum likelihood technique is developed. Finally, a method of image enhancement based on directional operators and template matching attempts to diminish noise while preserving edges. These techniques are applied both in the image domain and in the sinogram domain and then evaluated using a similarity measure against the original Shepp-Logan phantom.

Peanoscanning was used to obtain the pixels from an image by following a scan path described by a space-filling curve, the Peano-Hilbert curve. The Peanoscanned data were then compressed without loss of information by direct Huffrnan, arithmetic, and Lernpel-Ziv-Welch coding, as well as predictive and transform coding. In our implementation, tested on seven natural images, Peano-differential coding with an entropy coder gave the best results of reversible compression from 8 bits/pixel to about 5 bits/pixel, which was better than predictive coding of equivalent raster-scanned data. An efficient implementation of the Peanoscanning operation based on the symmetry exhibited by the Peano-Hilbert curve is also suggested.

Restoration is ultimately a problem of statistical estimation. An optimal filter is estimated to restore binary fax images. The filter is an approximation of the binary conditional expectation that minimizes the expected absolute error between the degraded image and the ideal image. It is implemented as a morphological hit-or-miss filter. Estimation methodology employs a model-based simulation of the degradation due to the fax process. Simulated images are used to generate data from which the filter is estimated. The methodology presented can be used for other classes of images. Restoration examples are given.

Distinct from gray-scale image enhancement, color image enhancement has two additional requirements. One is to preserve the natural color of the original image, and the other is to effectively present the information contained in the luminance and color components. We show that the color difference map and the luminance map represent different aspects of a color image. An enhancement scheme for color images is proposed based on this observation. In this method, an adaptive luminance masking function is generated from the local luminance contrast and color contrast. Overenhancement distortion, which is caused by the constraint of the display range, is also analyzed. A refinement algorithm is developed to solve this problem.

We describe a modification of the error-diffusion algorithm that produces binary output patterns having a clustered-dot appearance rather than the dispersed-dot appearance commonly associated with error diffusion. This is obtained by introducing a large dynamic range threshold function into the algorithm that is a function of both the spatial coordinate and input value.

This PDF file contains the editorial “Book Rvw: Computer and Robot Vision, " by Robert M. Haralick and Linda G. Shapiro for JEI Vol. 3 Issue 02