This book is a good reference for the Discrete Cosine Transform. The mathematics is only moderately difficult, and should provide little difficulty for a talented engineer graduate, say.
And yet I can't give this book a great review. It's all rather clinical, with little attempt to provide overview or insight. One just doesn't come away terribly enlightened.
The main reason provided as to why the DCT might be better than the DFT in some instances is that it is a more compact encoder of Markov sequences. Great, but what does this mean in practise? Well, the authors give many applications, but they are dealt with briefly, with little insight, and in a encyclopedic manner. Better would have been a few simple applications, done in more detail, where the authors made crystal clear why the DCT is the tool for the job.
I was hoping to come away with an in-depth understanding as to the properties of the DCT, where it's best applied, and why. I was disappointed.