Quality Augmentation of Metrics in Video Codec with Noise Removal and Discriminant Analysis

Abstract

Video compression is a technique for removing redundancy of data that helps in efficient storage and transmission of video which is essential for video streaming in current scenario. Many codecs and standards are available for video compression but the complexity that arises during encoding, decoding of video should be reduced for better performance. Methodology proposed here applies a machine learning technique to reduce the complexity along with a pre-processing method of noise removal on the frames. The choice in size of blocks during partitioning for frame prediction depends on a discriminant analysis classifier. Commonly used metrics like Peak Signal Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), Compression ratio, encoding time were analysed to check the image quality. Experimental results illustrate that splitting the blocks using the classifier can give time reduction by 32.68% with QCIF sequences and with no comprise in image quality.