Image Processing

<电子科技大学 格拉斯哥学院 刘宇>

Introduction

Image Processing

Image processing is the technique of using digital devices to analyse and operate the images in purpose of extract, recognization, combination or others. For the digital images, it refers to a two-dimension array for each pixels position and a single or multiple channel for the color, most frequently in grayscale, RGB or HSV, image processing is the process of extracts the value of each pixel and operation with these captured value.

Color system

In most of the case, image processing is based on three different color system: Gray, RGB and HSV.
Gray is based on monochrome images, from the grayscale value it can provide extra-color between black and white, this system is the most clear one in the operation, and it can also support the recognization and HSV processing.
RGB stands for Red, Green and Blue, means the images’ color is the mixture of these three channels. This color system is better for digital edition, but for human it is hard to understand the mixture result from this system, so in some case the RGB-HSV transfomation plays a main role in image processing.
HSV is short for Hue, Saturation and Value, this color system is also called Hexcone Model. This color system is the nearest one to the reality, when we are extracting characristic color we often use this system.
In addition, the other color system like CMY for printer are not usually used in image processing.

Transformation

Geometry transformation and mathematic transformation is the basic operation of image processing.
Geometry transformation, basically translation, rotation, scale, affine and projective. These are tools for image processing, it based on matrix operation and gives a posibility for image to be extracted and transformation for certain purpose.
Mathematic transformation like DFT and DCT is tool to achieve geometry transformation, and also with these tool we can recovery zipped image or generate infinate pattern, also it can support identification.

Application

Image recovery

For image zipped or broken, image processing can used to recovery these image. For zipped image, it need to be continus transformed. And for broken image, a method of process is that we can extract the certain color and analyse the array function of this color and its near area as a reference line, construct these reference net in characteristic color and relationship in chromatics.

Image combination

As for this process, the images need to be analysed to find the similar part. But in most of the case, due to enviroment and other reason, the angle of images are usually not the same, for this problem we need to execute geometry transformation and identify the willing result.

Identification of Objects

To achieve this goal, we first need to get enough base data, extract the similar part and do geometry transformation to transform the base image value into characteristic function group to describe the relationship between each part of the object to build a virtual 3-D data and in the target image it will be used by 3-D to 2-D objective to test the matching of the data object.