The paper presents a complete coder structure for compression of color still images represented with 24 bits per pixel, i.e. 8 bits for each of the primary colors red, green, and blue (RGB images). The RGB components of the original image is first transformed into one luminance component (Y) and two chrominance components (U, V). Each component is then coded in a coder consisting of a parallel filter bank for signal decomposition, uniform threshold quantization for irrelevancy reduction, and dynamic allocation of arithmetic coders for efficient redundancy reduction. The Y-component requires a significantly higher bit rate than the UV-components, which can be lowpass filtered and decimated. It is shown that the acceptable decimation factor to avoid color artifacts is highly dependent on the image to be coded. For high quality color image coding the chrominance decimation factor may be as low as 4. The coding results of the proposed scheme are compared to the JPEG coder both through objective and subjective image quality evaluation. Furthermore, the performance differences between the two coders are analyzed by separating the coding scheme into four parts: 1) signal decomposition, 2) quantization, 3) entropy coding, and 4) color coding.