Image processing process on image into ubiquitous format. The input image should be from camera CCD array, optical photo receptors and rays in virtual camera. This input is then processed and output is obtained from computer monitor. Most projects of image processing tries to unravel the problem that occur in its techniques. There are types of image processing concepts they are image measurement and image enhancement. Matlab simulation tools are used in color image processing. We offer a promising new topic on image processing such as image processing with embedded system.
Image processing process:
Objects that are obtained from sensing process are the main process of image processing. This process is executed through digital camera and sensors. After that feature extraction to be done. The features that are to be extracted are shape, color, geometric functions and texture. To make all this process possible various algorithms should be used. Such algorithms are as follows
- Nearest neighbor classifier.
- Artificial neural networks.
- Decision tree classifier.
Nearest neighbor classifier:
An object is classified by this k-nearest is the widely used classifier. There are three nodes present in a decision tree. Training and testing are the two types of process that enunciates classification. Only the features that are close are taken for classification. Various applications are benefited using machine learning algorithm. Such applications are speech recognition, filter spam, autonomous vehicles and applications which predict heart attack and stock prizes.
Artificial neural network:
It is a computational system. Structure processing method and learning ability of a biological form of brain is the base for this network. Input layer, hidden layer and output layer are the three types of layer.
These are four different learning rules such as error correction, competitive learning, hebbian and Boltzmann.
Decision tree classifier:
A systems item are classified using decision tree classifier. Root node, internal node and leaf node are three types of decision tree. There are no incoming edges in the root node but it has multiple output nodes. One input node and two or three output nodes are present in internal nodes. This algorithm brings out optimal output in fair share of time.
Many algorithms are needed in pattern recognition. It performs speech recognition, stock marked prediction, character recognition, weather prediction and medical diagnosis. There are three basic methodologies. They are linguistic, mathematical and heuristic
Projects on image processing related to pattern recognition are offered by us. They are by far the most advanced domain of image processing.