Still image compression with lattice quantization in wavelet domain

One of the most important directions in the area of coding of still images is based on the utilization of spatial correlation between neighboring pixels in an image. The correlation is developed by mapping the image data into a frequency plane using one of the transforms, the wavelet transforms being a significant option. The resulting set of transform coefficients represents the original image. The coefficients are subsequently quantized and coded to achieve the desired compression ratio. The quality of an image-compression system is assessed by the quality of the output images, the compression ratio, and the time required for encoding and decoding. An optimal bit-allocation routine is developed to minimize the distortion introduced by the procedure of quantizing the wavelet coefficients. The use of probability distribution function of coefficients significantly reduces the amount of computation required to obtain the solution of the distortion function, which in turn lowers the overall time of encoding.