This comprehensive reference on the major graphics file formats and the compression technologies they employ is an indispensable resource for graphics programmers, especially those developing graphical applications for the Web.
Compressed Image File Formats: JPEG, PNG, GIF, XBM, BMP examines the most common graphics file formats in detail and demonstrates how to encode and decode image files for each. In particular, this book offers in-depth coverage of the elaborate JPEG and newer PNG formats, providing clear explanations of complex concepts, experience-based practical techniques, and plentiful code examples. GIF, XBM, and BMP are also covered, with a focus on some of the less familiar and less well-documented features of these common file formats.
Specific topics covered include:
The accompanying CD-ROM contains the complete source code for all of the image formats covered in the book, as well as working examples and sample images. If you want to learn how to read and write graphic file formats for the Web--including PNG and JPEG files--there is no better resource than this book. System requirements: Windows 95 or Windows NT, Borland C++ Builder 3, or Microsoft Visual C++ 5.0.
The Representation of Images.
Vector and Bitmap Graphics.
True Color versus Palette.
Byte and Bit Ordering.
A Common Image Format.
Reading and Writing XBM Files.
JPEG Compression Modes.
What Part of JPEG Will Be Covered in This Book?
What are JPEG Files?
SPIFF File Format.
Interleaved and Noninterleaved Scans.
Huffman Coding Example.
Huffman Coding Using Code Lengths.
Huffman Coding in JPEG.
Limiting Code Lengths.
Decoding Huffman Codes.
DCT in One Dimension.
DCT in Two Dimensions.
Basic Matrix Operations.
Using the 2-D Forward DCT.
Decoding Data Units.
Processing DCT Coefficients.
Restart Marker Processing.
Overview of JPEG Decoding.
Output File Structure.
Doing the Encoding.
Data Unit Encoding.
Huffman Table Generation.
Factoring the DCT Matrix.
Scaled Integer Arithmetic.
Merging Quantization and the DCT.
Component Division in Progressive JPEG.
Processing Progressive JPEG Files.
Processing Progressive Scans.
MCUs in Progressive Scans.
Huffman Tables in Progressive Scans.
Data Unit Decoding.
Preparing to Create Progressive JPEG Files.
Encoding Progressive Scans.
Data Unit Encoding.
Compressed Data Format.
Color Representation in PNG.
Decompressing the Image Data.
Huffman Coding in Deflate.
Compressed Data Format.
Compressed Data Blocks.
Writing the Decompressed Data to the Image.
Deflate Compression Process.
Huffman Table Generation.
The purpose of this book is to instruct the reader on how to write software that can read and write files using various 2-D image formats. I wanted to write a book that explains the most frequently used file formats with enough depth for the reader to implement them, as opposed to one that covered many different formats at a high level or one that avoided the more difficult image formats. As a result, I chose to cover the image file formats that are associated with Web browsers. Those covered in this book (BMP, XBM, JPEG, GIF, and PNG) represent the vast majority of image files that can be found on the Internet. They employ a wide range of encoding techniques and range in implementation difficulty from simple to very complex.
The inspiration for this book was my own frustration resulting from the lack of information on how to implement encoders and decoders for the more complex file formats. Most of the information available was at too high a level, left major gaps, or was very difficult to decipher. I have tried to create a bridge between the programmer and the standards documents.
One issue I faced at the start of this project was which programming language to use for the examples. The intention was to create a book on graphics file formats rather than one on how to write programs to read and write graphics files in a particular language. Therefore, I debated using a language that is easy to read (e.g., Pascal or Ada) or the one most people are likely to use (C++). In the end I felt that its widespread use made C++ the best choice. To make the examples more understandable for non-C++ programmers, I have carefully avoided certain C++ language constructs (e.g., expressions with side effects and integer/boolean interchangeability) that would make the code difficult for them to understand.
In order to make the encoding and decoding processes as clear as possible, I have used a Pascal-like pseudo-code. C++ is used for complete function implementations and pseudo-code for illustrative fragments. These fragments generally contain no error checking.
Because of their generally large size, it was not possible to include working source code for the formats in the book itself. Instead, the accompanying CD-ROM contains the complete source code for encoders and decoders for almost all of the image formats covered. The reader should use the pseudo-code in the text to learn how processes work and the C++ examples on the CD to see how to implement them.
Generally, the decoders implement more features than the encoders. In the decoders I have implemented all of the features needed to decode files that a reader will have any likelihood of encountering on the Internet. For the sake of clarity, the encoders generally implement a smaller feature subset.
In writing the programming examples I have given clarity precedence over execution efficiency and instant portability. The source examples will compile, without modifications, on Microsoft Windows using both Borland C++Builder V3.0 and Microsoft Visual C++ V5.0. Other compilers generally require some modifications to the code.
The descriptions of the encoders and decoders for the various file formats frequently employ the term "user" to describe the source of certain input parameters to the encoding or decoding process. By this I mean the user of the encoder or decoder, not necessarily the person typing at the keyboard. Since image encoders and decoders are incorporated into other applications, such as image viewers and editors, the user in this case would most likely be another piece of software. However, in many situations the "user" application may get some of these parameters directly from a human.
Just as this is not intended to be a book on C++ programming, it is also not intended to be a book on programming in a specific environment. For that information readers will need a book for their particular system.
A project as large as producing a book requires the involvement of many people. Mike Bailey, Eric Haines, Tom Lane, Shawn Neely, and Glenn Randers-Pehrson reviewed the manuscript and provided many invaluable suggestions. Glenn also arranged for me to get the latest proposed PNG standards for the CD. My fellow aviator, Charlie Baumann, was kind enough to provide several of the photographs. Ralph Miano and Margaret Miano assisted with preparing the manuscript. Jean-Loup Gailley answered all my questions on ZLIB. Albert "The Chipster" Copper compiled examples on systems I did not have access to. Most important, Helen Goldstein at AWL guided the process from start to finish.John M. Miano