Image Processing Chapter 10
chapter Ten
10.1 Color description
The difference between color features and other visual features (emphasis)
The color feature is one of the basic features of an image.
Color feature is the most widely used visual feature in image retrieval and recognition. Compared with other visual features, it has weaker dependence on the size, direction and angle of view of the image, so it has higher stability .
The color description includes simple grayscale features and histogram features.
10.1.1 Simple gray-scale features (just understand)
Including gray average, gray maximum, minimum, median, order value and variance, etc.
10.1.2 Histogram characteristics (just understand)
The image grayscale histogram can be considered as an estimate of the grayscale probability density of the image. Including the following features: mean, variance, energy, entropy
10.2 Texture description
Repetitive structures such as cloth patterns, grass, and brick floors are called textures. Generally speaking, texture is a description of the spatial distribution pattern of pixel gray levels of an image , reflecting the texture of the item , such as roughness, smoothness, granularity, randomness, and standardization.
10.2.3 Gray-level co-occurrence matrix
Gray-level co-occurrence matrix method is one of the important methods to describe texture characteristics, it can more accurately reflect the texture roughness and repeating direction .
The picture above is a gray co-occurrence matrix with a distance of 1 at 0 degrees
(the positive direction of the x-axis is 0 degrees, and the degree increment is counterclockwise)
The co-occurrence matrix can reflect the main characteristics of the image texture. For flat areas, coarse texture areas, and pixels close to each other generally have similar gray levels. Therefore, when the distance d is small, the diagonal and nearby elements have large values and fine texture. The value of each element of the co-occurrence matrix of the region is relatively uniform.
10.2.4 Spectrum characteristics
The spectral feature is a texture description based on the Fourier spectrum. The global texture pattern is difficult to detect in the spatial domain, but it is easy to distinguish when converted to the frequency domain.
The spectral texture is very effective for distinguishing the difference between periodic mode or aperiodic mode and periodic mode. Generally, the global texture mode corresponds to the area where the energy is very concentrated in the Fourier spectrum, that is, the peak protrusion.
10.3 Boundary description
The boundary description mainly uses the external characteristics of the area, that is, the boundary of the area to describe the area.
10.3.1 Boundary expression
10.3.1.1 Chain Code
(1) Definition The
chain code is used to represent the boundary line composed of successively connected straight line segments with specified length and direction.
In a typical case, this representation method is based on 4 or 8 connections of line segments.
The direction of each segment is coded using a digital numbering method
. The picture above is an eight-way chain code
(2) Chain code normalization
Circulate in one direction to minimize the value of the natural numbers that they constitute, and use the starting point of the chain code after conversion as the starting point of the normalized chain code of this boundary. For example, the normalized chain code of the 4-way chain code 10103322 is 01033221
10.3.1.2 Polygons are similar
Use the polygon method based on shrinkage or aggregation or classification to determine the boundary
10.3.2 Boundary description features
Including boundary length, diameter, major axis, minor axis, eccentricity, curvature
10.4 Area description
10.4.1 Simple area description (area, center of gravity)
10.4.2 Topological description (holes, number of connections, Euler number = number of connections-number of holes)
10.4.3 Shape description (roundness, eccentricity)
10.4.3.1 Shape parameters
The roundness F, the circumference B, and the area A
can be seen from the above formula. When a continuous area is circular, F is 1, and when the area is of other shapes, F is greater than 1. That is, the value of F reaches the minimum when the area is a circle.
In addition to near circularity, there is also eccentricity: the ratio of the primary and secondary axis of the area
10.4.4 Moments
When a region R is only given in the form of its internal points, we can use the moment feature to describe it, which is invariant to changes in size, rotation and translation.