Emerging Trends in Image Processing, Computer Vision and Pattern Recognition

Leonidas Deligiannidis, Hamid R Arabnia

• Publisher: Morgan Kaufmann
• Year: 2014
• Print ISBN: 9780128020456
• eBook ISBN: 9780128020920
• Topic: Informatik
• Subtopic: Computer Vision & Mustererkennung

Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition discusses the latest in trends in imaging science which at its core consists of three intertwined computer science fields, namely: Image Processing, Computer Vision, and Pattern Recognition. There is significant renewed interest in each of these three fields fueled by Big Data and Data Analytic initiatives including but not limited to; applications as diverse as computational biology, biometrics, biomedical imaging, robotics, security, and knowledge engineering. These three core topics discussed here provide a solid introduction to image processing along with low-level processing techniques, computer vision fundamentals along with examples of applied applications and pattern recognition algorithms and methodologies that will be of value to the image processing and computer vision research communities.Drawing upon the knowledge of recognized experts with years of practical experience and discussing new and novel applications Editors' Leonidas Deligiannidis and Hamid Arabnia cover;- Many perspectives of image processing spanning from fundamental mathematical theory and sampling, to image representation and reconstruction, filtering in spatial and frequency domain, geometrical transformations, and image restoration and segmentation- Key application techniques in computer vision some of which are camera networks and vision, image feature extraction, face and gesture recognition and biometric authentication- Pattern recognition algorithms including but not limited to; Supervised and unsupervised classification algorithms, Ensemble learning algorithms, and parsing algorithms.- How to use image processing and visualization to analyze big data.- Discusses novel applications that can benefit from image processing, computer vision and pattern recognition such as computational biology, biometrics, biomedical imaging, robotics, security, and knowledge engineering.- Covers key application techniques in computer vision from fundamentals to mid to high level processing some of which are camera networks and vision, image feature extraction, face and gesture recognition and biometric authentication.- Presents a number of pattern recognition algorithms and methodologies including but not limited to; supervised and unsupervised classification algorithms, Ensemble learning algorithms, and parsing algorithms.- Explains how to use image processing and visualization to analyze big data.

Part 1

Image and Signal Processing

Chapter 1

Denoising camera data

Shape-adaptive noise reduction for color filter array image data

Tamara Seybold¹; Bernd Klässner²; Walter Stechele^{2 1} Arnold & Richter Cine Technik, München, Germany
² Technische Universität München, München, Germany

Abstract

While denoising readily processed images has been studied extensively, the reduction of camera noise in the camera raw data is still a challenging problem. Camera noise is signal-dependent and the raw data is a color filter array (CFA) image, which means the neighboring values are not of the same color and standard denoising methods cannot be used. In this paper, we propose a new method for efficient raw data denoising that is based on a shape-adaptive DCT (SA-DCT), which was originally proposed for non-CFA data. Our method consists of three steps: a luminance transformation of the Bayer data, determining an adequate neighborhood for denoising and hard thresholding in the SA-DCT domain. The SA-DCT is applied on realistic CFA data and accounts for the signal-dependent noise characteristic using a locally adaptive threshold and signal-dependent weights. We additionally present a computationally efficient solution to suppress flickering in video data. We evaluate the method quantitatively and visually using both realistically simulated test sequences and real camera data. Our method is compared to the state-of-the-art methods and achieves similar performance in terms of PSNR. In terms of visual quality, our method can reach more pleasant results compared to state-of-the-art methods, while the computational complexity is kept low.

Keywords

Color denoising

Camera raw data

Color filter array

CFA data

Implementation cost

Video denoising

1 Introduction

While denoising is an extensively studied task in signal processing research, most denoising methods are designed and evaluated using readily processed image data, e.g., the well-known Kodak data set [1]. The noise model is usually additive white Gaussian noise (AWGN). This kind of test data does not correspond nowadays to real-world image or video data taken with a digital camera.

To understand the difference, let us review the color image capturing via a digital camera, which is the usual way of image capturing nowadays. One pixel captures the light intensity, thus the sensor data corresponds linearly to the lightness at the pixel position. To capture color data, a color filter array (CFA) is used, which covers the pixels with a filter layer. Thus the output of the sensor is a value that represents the light intensity for one color band at one pixel position. This sensor data cannot be displayed before the following steps are applied: the white balance, the demosaicking, which leads to a full color image and the color transformations, which adapt the linear data to displayable monitor data adapted to the monitor gamma and color space. These steps lead to a noise characteristic that is fundamentally different from the usually assumed AWGN: through demosaicking it is spatially and chromatically correlated and through the nonlinear color transformations the noise distribution is unknown.

As this noise characteristic cannot be easily incorporated into the denoising methods, we propose to apply the denoising to the raw CFA data—the mosaicked data linear to the light intensity with uncorrupted noise characteristics. In the raw data we observe noise with a known distribution and a signal-dependent variance, which can be precisely estimated based on measurements [2]. However, despite the richness of denoising methods, denoising color image raw data has been less studied until now. Hirakawa extended a wavelet-based method to CFA data [3]. Zhang proposed a principle component analysis (PCA) based solution [4]. The state-of-th...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• Contributors
• Acknowledgments
• Preface
• Introduction
• Part 1: Image and Signal Processing
• Part 2: Computer Vision and Recognition Systems
• Part 3: Registration, Matching, and Pattern Recognition
• Index