Segmentation is the underlying concept for creating objects from pixels. The segmentation process involves dividing the image into regions or objects have common properties. Typical image segmentation techniques involve one of the two process:

1. Region merging according to some measure of homogeneity

2. Separation of objects by finding edges using the gradient of digital numbers (DNs) between neighboring pixels.

Region- margining approaches can be divided into two approaches:

1. Region Growing

2. Region split and merging

Region growing technique involves the pixel aggregation and starts with a set of seed points. From these seed pixels, the regions are grown by merging neighboring pixels that have similar properties.

Image segmentation by pixel aggregation:

It is based on the concept of region without needing the seeds to start the segmentation process. At the beginning of the algorithm, each pixel has its own label. A pixel is included in a region if it is four-connected or eight –connected to this region and has a color value in the region, the region’s mean color is updated. For this updating, recurrent formulas are used. One or two simple scans of the color image are applied: One pass from the left to the right and from the top to bottom and can be followed by additionally during each scan. The pixel aggregation process results in a set of regions characterized by their mean colors, their sizes and lists of pixels that belong to proper regions. The regions are generated sequentially. The basic version of the algorithm works in the RGB color space. During the tests, it was observed that if the value of the parameter the increased, the number of the regions R in the segmented image simultaneously decreased. The results for the color image flowers are shown in below images. Too low a value of the parameter d increased which leads to over segmentation and too high a value causes under segmentation.