Superficial Understanding of Image Processing

Writer Information

姓名：熊汇雨
学校信息：电子科技大学格拉斯哥学院
班级：EEE6
学号：2017200506007

This article is for me after the 2017 freshmen seminar of Glasgow College, University of Electronic Science and Technology, collating the relevant data, and put forward some of my own immature views. We welcome your opinions and suggestions on the shortcomings. If there are any errors, we sincerely hope that readers will criticize and correct them.

Preface

After listening to the freshman seminar in the first semester, I found that image processing is widely used in our daily life, such as staellite image processing, medical image processing, face recognition, feature recognize and vehicle obstacle recognition…………Therefore, how to recognize and process images more accurately is a hot topic. In this course, the teacher who majors in computer science gave us a simple introduction about the aspect. I began to have a general impression of the technology.

Preparation

Image processing is a technical term in the field of computer science. It has the following professional explanations in Wikipedia:

In computer science, digital image processing is the use of computer algorithms to perform image processing on digital images. As a subcategory or field of digital signal processing, digital image processing has many advantages over analog image processing. It allows a much wider range of algorithms to be applied to the input data and can avoid problems such as the build-up of noise and signal distortion during processing. Since images are defined over two dimensions (perhaps more) digital image processing may be modeled in the form of multidimensional systems.
link.

I have always been interested in pictorial forms. So at the end of the first semester, I will think of using some knowledge of image processing to complete my final paper about linear algebra.
Combining the basic knowledge of matrix in linear algebra class and the information learned from all aspects, I initially determined the singular value decomposition(SVD) to solve this simple problem of pixel blurring.

Time Planning

Paper

First, after observing the actual problem of pixel blurring, I asked the following question:

• How to keep the image sharpness after compression

By querying relevant information, I learned that singular value decomposition can solve this problem.

Singular value decomposition (SVD) is a matrix transformation method based on eigenvectors. It has been applied in signal processing, pattern recognition, and digital watermarking technology. As the image has matrix structure, it is proposed to apply singular value decomposition to image compression, and has been successful.
link

Analysis

In the operation of the image, because the dimension of the matrix is generally large, the computation of the singular value decomposition directly is large. Then the image can be decomposed into sub blocks. Decompose their singular values, determine their number, and reconstruct each sub block. In addition to the easier computation of singular value decomposition for smaller matrices, the other aim is to make use of the inhomogeneous complexity of the original image. If a certain part of the image is relatively simple, only a small number of singular values are needed to achieve a satisfactory approximation effect.
In order to ensure the quality of images, more singular values are needed. However, the number and size of each sub block are different, so it is possible to choose the appropriate number of singular values adaptively for each block.

Advantages and Disadvantages

The image obtained by singular value decomposition will occupy smaller storage space compared with the original image, and the definition will not change much. Therefore, this method is of great benefit to the development of modern science and technology.
But the operation of singular value decomposition transform is more complicated. Whether we can find a fast SVD transform algorithm still need to be considered and improved.

Postscript

Trough freshman seminar in the first semester, I learned that the pixels can be the matrix, which makes the image clear. Thank our college for the popularization of knowledge. It provided the foundation for academic researches.

Bibliography

[1] Zhou Long, Huang Lingxiao, Mou Yi, “Application of singular value decomposition in image compression of stored grain pests”. Journal of Huazhong University of Science and Technology: Natural Science Edition , 2009 (1) :46-48

It is written by Xiong huiyu(熊汇雨) from EEE6(2017200506007)