Biomedical Image Analysis

Detection of Circles and Ellipses With The Hough Transform (Biomedical Image Analysis)

Many shapes in medical images can be approximated by circles or ellipses. The principal idea of the Hough transform, the accumulation of votes in parameter space, can be applied to both circles and ellipses, since both shapes can be described analytically. A circle is described by three parameters: radius r and the two center coordinates, […]

Generalized Hough Transform (Biomedical Image Analysis)

The idea of the Hough transform, the accumulation of votes, can be extended to detect arbitrary (i.e., nonanalytical) shapes.2 For this purpose, the shape template is decomposed into straight edge sections. In the extreme case, each edge pixel may represent one edge section. An arbitrary reference point is chosen. A frequent choice is the centroid […]

Randomized Hough Transform (Biomedical Image Analysis)

The inherent disadvantages of the Hough transform, that is, large memory space use and the challenge of finding meaningful local maxima in Hough space, led to a different approach to curve parametrization: the randomized Hough transform.21 Strictly, the randomized Hough transform is not a transform since the original space cannot be fully reconstructed, due to […]

Biomedical Examples (Biomedical Image Analysis)

Although development of the Hough transform is driven primarily by computer vision applications, the Hough transform is frequently used in biomedical image analysis. Medical images pose additional challenges to image analysis, primarily noise and natural shape variation. For this reason, most medical applications of the Hough transform include additional image processing steps or variations of […]

Texture Analysis (Biomedical Image Analysis)

Each image feature contains three major elements of information. For each feature, this is the average image value (e.g., intensity, exposure, density), the shape, and any systematic local variation of the image values. This local variation is called texture. Some examples that can be found in the UIUC texture database1,26 are shown in Figure 8.1. […]

Statistical Texture Classification (Biomedical Image Analysis)

The intensity distribution of pixels in an image or a region of an image is characterized in the intensity histogram. The number of pixels is plotted as a function of their intensity, providing a discrete function n(I). For the purpose of texture classification, with the factor N being the total number of pixels. The normalized […]

Texture Classification With Local Neighborhood Methods part 1 (Biomedical Image Analysis)

Texture Classification Based on the Co-occurrence Matrix Many texture classification operations require computation of the co-occurrence matrix. These methods were pioneered in 1973 by Haralick et al.,15,16 who formulated 14 texture metrics that build on the co-occurrence matrix. The gray-scale cooccurrence matrix C00 (ij) is a two-dimensional histogram of gray values i and j where […]

Texture Classification With Local Neighborhood Methods part 2 (Biomedical Image Analysis)

Laws’ Texture Energy Metrics K. I. Laws proposed a different approach to texture classification25 by suggesting five different one-dimensional convolution kernels: Although the kernel size may be larger or smaller in different implementations, the length 5 kernels are being used most commonly. L5 is a Gaussian-type blurring kernel that provides the smoothed gray level (L) […]

Frequency-Domain Methods For Texture Classification (Biomedical Image Analysis)

In the frequency domain, texture properties such as coarseness, graininess, or repeating patterns can be identified. Of primary interest is the spectrum’s magnitude or the squared magnitude, (i.e., the power). Since the Fourier transform does not retain spatial information, only global properties can be examined. To illustrate the relationship between the visual appearance of a […]

Run Lengths (Biomedical Image Analysis)

The run-length method13 to classify texture is very popular and widespread in the analysis of biomedical images. A run is a sequence, in a straight scan direction, of pixels with identical image value. The associated run length is the length of the run, usually the number of pixels for the horizontal or vertical scan direction, […]

← Previous Entries

Next Entries →