ABSTRACT
Thresholding is a popular, fast, and computationally inexpensive segmentation technique. If the objects in an image are disjoint and their gray levels are clearly distinct from the background, thresholding is the appropriate type of segmentation. In a text, the characters are generally darker than the paper. In pictures of mitotic cells, chromosomes or tracks are darker or lighter than the background. In all these cases, the gray-level histogram of the picture displays the peaks corresponding to the two gray levels: characters and the paper, chromosomes or tracks and background. In thresholding, an appropriate threshold is chosen that separates these peaks and segments the picture into two or more segments: one (or more) for the object(s) and the other for the background. A histogram containing a single peak is called unimodal, two peaks is called bimodal, and multiple peaks is called multimodal. A complete segmentation of an image R is a fi nite set of regions R1, R2, R3, …, RN such that
1 and
R R R R i j =
= ∩ = φ∪ ≠
Thresholding is a transformation of an input image A into a segmented output image B as follows:
bij = 1, for aij ≥ T = 0 for aij ≤ T, where T is the threshold
bij = 1, for the image pixels that belong to the object class bij = 0, for the image pixels that belong to the background class
However, the selection of the threshold is very crucial in segmenting an image. If the threshold is not selected properly, proper segmentation cannot be obtained. One such segmentation performed using threshold values too high and too low is shown in Figure 5.1. If the threshold value is more than the actual threshold, the image is oversegmented and if the threshold value is less than the actual threshold, the image is less segmented.
