Embedding Parallelism In Image Processing Techniques And Its Applications

Abstract

Image processing become demanding and attracting research domain due to its versatile application in a different field such as military field, medical imaging, authentication of the digitization process through signature recognition, face recognition, the agricultural field, etc. Due to improved technology and massive development in cost-effective image acquisition equipment, the size and number of images increase day by day. In some domains, such as medical, military applications, accuracy, as well as real-time image analysis, is the most important criterion for evaluating the performance of the system. As each application is having its own requirement, every system demands real-time, accurate, less expensive, and more extensive computation. Most of the application of computer vision, image processing, pattern recognition deploys the feature extraction to use the machine learning and deep learning methodology. All these techniques and architecture requires huge time to extract the features of the images which could not be acceptable in some of the real-world application. To reduce the time and make the computation faster and efficient, the concept of parallel processing is embedded in image analysis. This study aims to provide an introduction as well as the need of parallelism in image processing. The mostly used parallel processing using CUDA and GPU is briefly discussed with their shortcomings. 234Finally, how parallel computation is used in machine learning and deep learning architecture for medical image analysis is discussed in detail to show the impact of parallel processing. Also, one case study based on brain tumor segmentation is elaborated in this chapter. The main purpose of this chapter is to summarize existing parallel image processing techniques and tools via various research and analysis and their limitation.