Color images have the potential to convey more information than a monochrome or black and white images. Each pixel for a true color image is stored as values of red, green, and blue. However, the RGB color model is not an efficient representation for compression task because there is significant correlation between RGB color components. Therefore, for compression, luminance-chrominance representations (such as YIQ) is implemented. A common approach to the color image compression was started by transform the RGB image to a desire color model, then applying compression techniques, and finally retransform the results back to RGB model. In this work, a color image compression scheme combining the wavelet transform and a modified vector quantization (MVQ) method is proposed. In wavelet transform, the low and high Haar filters are composed to construct four 2- dimensional filters, such filters are applied directly to the YIQ image to speed up the implementation of the Haar wavelet transform. Haar wavelet transform was used to map it to frequency bands. Bit allocation process and scalar quantization are implemented on the approximation subband while modified vector quantization mechanism is employed to encode other higher frequency subbands using small block size (so as decrease the codebook size) as the subband number increases. Since the encoding process is much easier when the range of coded parameters are positive, thus the coefficients values of codebook are mapped to the positive range. Finally S-Shift encoding process is performed. The analysis results have indicated that the proposed method offers a comparission performance up to (29/1) with little effects will be noticed on the image quality.