A Radial Basis Function Neural Network with Adaptive Structure via Particle Swarm Optimization
Tsung-Ying Sun, Chan-Cheng Liu, Chun‐Ling Lin, Sheng-Ta Hsieh, Cheng-Sen Huang · InTech eBooks · 2009
Necessity of video compression and standardsA low bit rate video coding (bit rate less than 64 kbs) needs high compression ratio (above 150).In high compression ratio video coding, block based coders introduce blocking artifact and ringing effect (Due to Gibbs phenomena) in the reconstructed signal.High compression image coding has triggered strong interests in recent years.In this type of coding, visible distortions of the original image are accepted in order to obtain very high compression factors.High compression image coders can be split into three distinct groups.The first group is called waveform coding and consists of transform and subband coding.The second group called second generation techniques, consisting of techniques attempting to describe an image in terms of visually meaningful primitives (contour and texture, for example).The third group is based on the fractal theory.An uncompressed video sequence for very low bit rate applications typically requires a bit stream of up to 10 Mbit/s.In order to achieve very low data rates, compression ratios of about 1000 : 1 are required to meet the needs of the large public.Intensive research has been performed in the last decade to attain this objective [Touradjebrahimi et. al, 1995].Variations of the recommendation H.261 for very low bit rate applications have been defined as simulation models.For these simulation models, severe blocking artifacts occur at very low data rates.Wavelet based video coding is developing a new area in video coding for last two decades.Because of the multi resolution property, wavelet tool is suitable for image enhancement and compression.Rather than a complete transformation into the frequency domain, as in DCT or FFT (Fast Fourier Transform), the wavelet transform produces coefficient values www.intechopen.comRecent Advances on Video Coding 290 which represent both time and frequency information.The hybrid spatial-frequency representation of the wavelet coefficients allows for analysis based on both spatial position and spatial frequency content.While most wavelet-based compression techniques employ the traditional critically sampled discrete wavelet transform (DWT), alternative wavelet transforms have recently been proposed.Specifically, the complex dual-tree discrete wavelet transform (DTCWT) has undergone investigation in 3D video-coding systems [B.Wang, et.al, 2004]. Particle swarm optimizationParticle Swarm Optimization (PSO) is global optimization technique based on swarm intelligence.It simulates the behavior of bird flocking [Kennedy et. al, 1995].It is widely accepted and focused by researchers due to its profound intelligence and simple algorithm structure.Currently PSO has been implemented in a wide range of research areas such as functional optimization, pattern recognition, neural network training and fuzzy system control etc., and is successful.In PSO, each potential solution is considered as one particle.The system is initialized with a population of random solutions (particles)and searches for optima (global best particle), according to some fitness function, by updating particles over generations; that is, particles "fly" through the N-dimensional problem search space to find the best solution by following the current better-performing particle.When compared to Genetic Algorithm, PSO has very few parameters to adjust and easy to implement.The variants of PSO's such as Binary PSO, Hybrid PSO, Adaptive PSO and Dissipative PSO are used in various image processing applications.Recently PSO has been extended to deal with multiple objective optimization problems [K. U. Parsopoulos et. al, 2002].In the past few years many research works have been focused on modifying PSO to handle multiple objective optimization problems known as multi objective particle swarm optimizer MOPSO.The fixed population size MOPSO and variable population size PSO (Dynamic PSO) are used throughout the evolution process to explore the search space to discover the non dominated individuals (particles).Most of the real life problems are multi objective nature.Multi objective optimization using PSO has been used in Digital image processing like image segmentation, data clustering etc.Here Video compression also viewed as a muliobjective one.The constraints are Means Square Error (MSE), Computation Time, and Computation complexity and compression ratio.In this chapter the only three constrains (Means Square Error (MSE), Computation Time and compression ratio) are considered for the PSO based optimization.All the three are minimization functions.The fixed population size MOPSO is used throughout the evolution process to explore the search space to discover the non dominated individuals (particles).First the image is decomposed into subband using the Dual tree wavelet transform and the subband coefficients are minimized using noise shaping method.After that the MOPSO algorithm is used to select the optimum subband which provides less mean square error and desired bit rate.In this MOPSO weighted average approach is used.The constraints total weight age is one.The obtained results are compared with the standard algorithms. Shortcomings in the conventional discrete wavelet transformTheoretically-sampled form of the wavelet transform (Discrete Wavelet transform DWT) provides the most compact representation; DWT has the following advantages: www.intechopen.com How to referenceIn order to correctly reference this scholarly work, feel free to copy and paste the following: M. Thamarai and R. Shanmugalakshmi