In this paper, we present a new wavelet image encoder that focuses on overcoming the main drawbacks of other wavelet
based-image coders; their high memory requirements and complexity. In our proposal, we tackle these problems in both
parts of the wavelet coder, both the DWT algorithm and the coding system. For the DWT, we propose the use of a line-based
algorithm, in which we get rid of the wavelet coefficients as soon as they have been calculated. Some previous
line-based proposals cannot be easily implemented, and here we describe a recursive algorithm that is implemented more
easily. For the coding system, we use a tree-based coder. These coders have evidenced good Rate/Distortion (R/D)
performance. However, it has not been clearly shown their excellent possibilities for fast processing of coefficients.
Previous tree-based proposals need the entire image in memory to work and so they cannot be used in our system. Other
wavelet encoders achieve good R/D performance, but exhibit high complexity. Our proposal is simpler and therefore
faster. Numerical results show that the compression performance of our tree-based encoder is equal or better than state-of-
the-art coders, such as SPIHT and JPEG 2000, except for highly detailed images. However, the real benefits of our
proposal is shown in the amount of memory required, which is reduced drastically (in the order of 25 times less memory
than SPIHT and more than 35 times less than JPEG 2000), and in its lower execution time (about four times lower than
SPIHT, and more than 20 times lower than JPEG 2000). These results show that our encoder is a good candidate for
many embedded systems and other memory-constrained environments, such as digital cameras and PDAs.