Before a file can be compressed using wavelets, we must first turn that file into a signal. The Codestream Syntax provides us with a method for such a conversion. First, I must note that JPEG 2000 compression is a tile-based scheme. Just as we learned in class, the larger, original image is first decomposed into multiple, smaller images called tiles. The dimensions of these tiles are always exact powers of two, and the tiles can be thought of as independent images.
Some of the most crucial components of data input stream, or signal, are the marker segments. There are two main types of marker segments: the main header marker segment and the tile-part header marker segment. The main header marker segment contains information on almost 25 different properties of the signal, including the start of the code stream, the image and tile size, the region of interest, and the quantization component. The tile-part header marker segment notes the start of the tile, the start of the tile data signal, the region of interest and quantization component of that particular tile, and many other parameters. We should note the importance of the tile-part headers. With the tile-part headers, the JPEG 2000 compression scheme is aware of each individual tile. Thus, the user can apply different compression tactics to each, individual tile. That means each tile can utilize different coefficients in quantization, each tile can be decomposed to different levels, and each tile can be accessed and referenced independently of each other. This last fact means a user can uncompress any portion of the compressed image. Suppose one doctor working in the Western Hemisphere wants to send the image of an x-ray to another doctor in the Eastern Hemisphere. With the current JPEG standard, the doctor must send the entire image, which could take a long time. With JPEG 2000, the doctor can send all of the image or just the interesting part of the image. This shows the importance of the tile-part header marker in the code stream.
