Industrial Machine Vision

Hue

the colour’s basic shade, measured around a colour wheel from 0° to 360°

Saturation

the purity/intensity of a colour measured from 0 (grey), 1 is fully vivid

Value

the apparent brightness of a colour measured from 0 (black) to 1 (max brightness)

Lightness

the perceived light/dark level of a colour measured from 0 (black) to 1 (white)

Chroma

the colourfulness of a colour compared with a grey of similar brightness measured from 0 to 1

Explain how brightness and contrast can be controlled by changing the histogram

Brightness is controlled by a histogram slide, by adding an offset to each grey level, which is done by changing the y intercept of the linear grey level mapping. Contrast is controlled by a histogram scale, where the gradient itself is tilted about a given setpoint in the linear grey level mapping, usually the histogram mean or median

Isodata thresholding technique

Isodata finds a threshold T that splits the histogram into two classes (background and foreground) by iteratively updating T until it converges.

T_new = (µ₁ + µ₂) / 2

Where µ₁ = mean of all pixels with grey level ≤ T, and µ₂ = mean of all pixels with grey level > T

Otsu thresholding method

Otsu auto thresholding is a global thresholding technique that tests every possible grey-level threshold and splits the histogram into two classes, foreground and background based on threshold T. It chooses the threshold T that minimises weighted within-class variance, equivalently maximising between-class variance.

v_w(g) = [s₁(g) · v₁(g)] / S  +  [s₂(g) · v₂(g)] / S

where  are the class probabilities and  are the class variances. The threshold with the smallest  is selected.

What is texture

Visual patterns or spatial variations in pixel intensity, which can be used to classify regions of perceived similar texture within an image. It is useful in industrial vision systems such as defect detection, medical imaging and remote sensing