Spatial or Block Coding

Spatial coding exploit the redundancy within the same frame. Each frame is decomposed into blocks of 8x8 pixels, where pixel values range from 0 to 255 (8 bits per pixel). These values are shifted to -127 and -128 (centered around zero). After that, Discrete Cosine Transform (DCT) is used to map the spatial data to the frequency domain (64 coefficients). The (0,0) coefficient represents the DC value. The DCT coefficients are then quantized to reduce their spread. The aim of this process is to zero out the higher frequencies. After that, a block-to-vector conversion is used in preparation for the entropy coding. This conversion is done by a zigzag pattern to scan the block to create a 64-element vector. After that, Run Length Coding (RCL) is used to replace the sequence of same consecutive bytes with the number of occurrences. Then a ``variation-zero suppression'' process is used to encode long binary bit strings containing mostly zeros.