Affiliations: [a] Department of Computer Science & Engineering, National Institute of Technology, Silchar, Assam, India
| [b] Department of Electrical Engineering, National Institute of Technology, Silchar, Assam, India
| [c] Department of Computer Science & Engineering, NERIST, Nirjuli, Arunachal Pradesh, India
Correspondence:
[*]
Corresponding author. Hussain Ahmed Choudhury, Department of Computer Science & Engineering, National Institute of Technology, Silchar, Assam, India. E-mail: hussain@rs.cse.student.nits.ac.in.
Abstract: Video compression is applied for reducing the requirement of hardware, bandwidth, hard drives and power consumption for storing and processing an excessive amount of data generated by videos. The computationally intensive and most time-consuming segment of video compression is known as motion estimation (ME). ME process can be regarded as an optimization problem where search is carried out in a predefined search area of the target frame to locate the identical macroblock (MB) corresponding to each MB in anchor frame by minimizing the objective function cum search criterion as minimum value of search criterion identifies the location of the best matching MB. Since the efficiency of ME decides the efficiency of Video compression, a rich number of fast block matching algorithms (BMAs) were reported to maintain the tradeoff between the computational complexity and visual experience of video during the ME process. Investigation reveals that most of the pattern-based BMAs are prone to the local optimum and stuck in sub-optimal results. Due to the emergence of various nature-inspired algorithms (NIA) like particle swarm optimization (PSO), genetic algorithm (GA), evolutionary algorithm, etc. and their application in optimizing all types of day to day problems has opened a new era in the field of ME. Our investigation focuses on the application of all types of NIA reported to date for optimizing the ME process in terms of speed, accuracy, and quality. This investigation will analyze all the NIAs and their methodologies through an extensive study of their accompanying publications and will enable us to do a detailed comparison to highlight the competitive advantage of soft computing techniques over existing pattern-based algorithms.
