INTRODUCING MACHINE LEARNING TECHNIQUE FOR JPEG COMPRESSION

The traditional compression technique and video coders such as JPEG, H.261, MPEG-1, MPEG-2 suffer from annoying blocking artifacts when they applying low bit rate coding .it can be occur because the inter correlation between the block is lost by block based prediction, quantization, transformation etc The commonly used JPEG image compression technique have good compression ratio but it also introduce blocking effects, to overcome the blocking artifacts we implement the machine learning technique to predict the DCT coefficient value and the pixel value of the compressed image. The result of our method gives the decreasing of the mean square error between our predicted image and original uncompressed image and it also gives the clear reduction of the blocking artifacts.