Schowengerdt, R.A., Techniques for Image Processing and Classification in Remote Sensing, Academic Press, Inc., Orlando, Florida, 1983.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....image into distinct classes in such a way that two pixels in the same class are spectrally similar. A map of distinct classes is a data product with a number of applications, including quicklook generation, data compression [11] image restoration [15] remote sensing change detection [14], and clutter reduction for weak signal detection [5] The standard algorithm is quite straightforward. Each iteration consists of a single pass over every pixel in the data set; during this pass, distances are computed from each pixel to each of the centers. The pixel is assigned to the cluster ....

R. A. Schowengerdt. Techniques for Image Processing and Classification in Remote Sensing. Academic Press, Orlando, 1983.

....individual pixel on the basis of its immediate neighborhood labels. The simplest method is the logical smoothing algorithm [4] which consists in assigning to each pixel the label of its 8 neighbors if all of them are similarly labeled. A slight modification of this method is the majority filter [5] assigning every pixel to the majority class within a window surrounding the pixel. One common limitation of these two contextual information fusion approaches is due to the fact that no interaction (or feedback) exists between input features and classification results. In fact, the first approach ....

R.Schowengerdt, Techniques for Image Processing and Classification in Remote Sensing, Academic Press, 1983.

.... also provides an organization of the data that can be useful for further downstream processing [7] Several authors have shown that clustering the data beforehand increases the performance of algorithms which attempt to learn features from a small number of examples [2, 6] Schowengerdt [9] suggests the use of image segmentation for change detection: a change in the segmentation is more likely to indicate an actual change on the ground, since the segmentation is relatively robust to changes in sensor performance and atmospheric conditions. It has also been demonstrated that the ....

R. A. Schowengerdt. Techniques for Image Processing and Classification in Remote Sensing. Academic Press, Orlando, 1983.

....of color. The use of multiple sensors is also not uncommon in the field of computational vision. Two or more similar sensors are required in stereo vision applications to estimate the depth of objects in the scene [2] Multiple imaging sensors are widely used in remote sensing applications [58]. Other applications of multisensor image fusion include navigation guidance [24, 39, 55, 72] medical imaging [74] object detection [56] and recognition [22] Image fusion is also used for compression of multisensor (or hyperspectral) images. The fused image can be stored or transmitted at a ....

R. A. Schowengerdt. Techniques for Image Processing and Classification in Remote Sensing. Academic Press, New York, 1983.

....of this paper is to offer simple examples of the poor performance of the correlation coefficient for image comparison, particularly with security applications in mind. We know of no similar presentation. The problems and limitations of the correlation coefficient have been discussed previously [2, 3, 13 17], but briefly and abstractly, without specific image examples. Furthermore, the knowledge that the correlation coefficient often performs poorly does not seem to have been communicated well to non experts. We conducted a survey of 50 image processing texts, introductory through advanced, that ....

....an undefined value to default to r=0 or r=1. Other problems with the correlation coefficient include possible bias [16] complexities of interpretation [1] over sensitivity to pixel noise and gain variations [2, 5, 15] difficulties in dealing with perspective or with moving illumination sources [2,15, 17], undesirable behavior for images containing too much fine structure or too little [2, 15] and trouble in dealing with images having strong spatial disparity gradients [2] There is an additional problem with the correlation coefficient that does not appear to have been discussed or demonstrated ....

R.A. Schowengerdt, Techniques for Image Processing and Classification in Remote Sensing, Academic Press, New York (1983), pp. 27-35, 47.

....have found that once clustering has been performed, the original data is no longer needed. Each pixel in an image is commonly categorized according to its spectral signature. Many methods are used for classifying multispectral data, including both supervised and unsupervised classification methods [1, 2]. When using supervised methods for data classification, a user selects training areas representative of several types of land cover, and a classifier is developed to discriminate between different classes. This classifier is then used to categorize the remaining pixels in the scene. Numerous ....

Robert A. Schowengerdt. Techniques for Image Processing and Classification in Remote Sensing. Academic Press, New York, New York, 1983.

....#30 cm Table 1.1: Principal divisions of the electromagnetic spectrum (Campbell [19] corrected) wavelength ( m) spectral reflectance 0.5 0.6 0.7 1.0 1.1 0.9 0.8 vegetation soil water Figure 1. 1: Generalised spectral reflectance curves for three di#erent materials (adopted from Schowengerdt [98]) 4 Chapter 1: Introduction detectors direction along track scan mirror spectrometer direction across track field of view instantaneous Figure 1.2: A multispectral mechanical scanner uses a scan mirror to direct the radiation inside the instantaneous field of view towards a spectrometer. ....

R.A. Schowengerdt. Techniques for Image Processing and Classification in Remote Sensing. Academic Press, Inc., 1983.

....to image analysis. The segmentation approach tries to group pixels of an image into segments. The classification approach classifies all the pixels in the image according to their spectral properties into several predefined classes representing specific features of certain ground cover types, see [Sch83]. 1.4.1 Image segmentation The proces of grouping pixels in an image is called segmentation. The idea is to form segments of pixels which correspond to objects seen in the original image. Segmentation methods can be grouped in 3 categories: ffl Edge finding ffl Region growing ffl Hybrid ....

....because region growing may not pass polygon borders. Sch95] gives a detailed account of advantages and disadvantages of using his hybrid segmentation method. 1.4.2 Image classification or clustering Image classification is a process in which pixels are catagorized by their spectral values. [Sch83] defined classification as the decision making technique which decides how to catagorize a pixel from its spectral features (values) The goal of classification therefore is to find areas in the image which have a similar spectral signature. Image classification is also often called clustering. ....

[Article contains additional citation context not shown here]

R.A. Schowengerdt. Techniques for Image Processing and Classification in Remote Sensing. Academic Press, 1983.

No context found.

Schowengerdt, R.A., Techniques for Image Processing and Classification in Remote Sensing, Academic Press, Inc., Orlando, Florida, 1983.

No context found.

R. A. Schowengerdt, Techniques for Image Processing and Classification in Remote Sensing, Academic Press, Orlando, 1983.

No context found.

R.A.Schowengerdt,"Techniques for Image Processing and Classification in Remote Sensing", Academic Press, 1983.

No context found.

R.A.Schowengerdt," Techniques for Image Processing and Classification in Remote Sensing",Academic Press, 1983.

No context found.

R. A. Schowengerdt, Techniques for Image Processing and Classification in Remote Sensing, Academic Press, Orlando, 1983.

No context found.

R. A. Schowengerdt, Techniques for Image Processing and Classification in Remote Sensing, Academic Press, Orlando, 1983.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC