13-03-2014, 03:57 PM
Edge Detection Using Orientation Based Similarity and Immunological Principles
ABSTRACT
Edge detection is one of the pre-processing operations performed during image analysis. Orientation-based Similarity (OBS) is a new approach for edge detection that is totally independent of intensity gradient and adaptive threshold. In this technique, the similarity among the pixels is obtained by comparing the Environment Mode (EM) of pixels. Initially a series of overlapping curves are formed by using OBS and then the pixels corresponding to the real edges are identified using the majority rule. The antigens and antibodies of the Artificial Immune System (AIS) are analogically mapped as real edges and the organized curves respectively. The technique uses the principle of imperfect detection wherein there is no exact match between the antibody and the antigen, thus improving the flexibility of the edge detection.
Overview of Edge Detection
An edge is the boundary between an object and its background indicating the presence of overlapping objects.
The intensity of an image moves from a low value to a high value or vice versa in the locations containing edges.
Conventional Edge Detectors
1) Sobel Edge Detector:
Sobel’s Edge Detection method is based on first derivative analysis
It is used to find the approximate gradient magnitude at each point in an input grayscale image.
Advantages:
Sobel operator uses the first derivative which is easy to calculate.
It calculates both edges and their orientation
Faster in computing the edges.
Sobel is one among the low cost algorithms and provides better results for uncompressed images.
Orientation based Similarity (OBS)
In the conventional edge detectors, it is necessary to define a final intensity threshold to identify the real edges.
OBS is a new perspective for edge detection that is absolutely independent from intensity gradient and adaptive threshold.
According to this technique, an edge is a series of pixel sewed by an instinct thread called OBS.
Similarity is the comparison between the Environment Mode(EM) of pixels present in the same edge.
The EM of every pixels are constructed by two different elements representing the two objects separated by the edge.
Orientation serves as the identity card for every edge.
OBS based Edge Detection using AIS
The real edges are considered as antigens, and the curves organized correspond to the antibodies.
Imperfect detection not only improves the curves condition but also makes the immune system more flexible.
Thus the curve forming algorithm is modified by using imperfect detection principle.
The memory cells that are produced during the detection process serve as the duplication of the previous curves and help in combining two different curves together to perform a continuous edge.