We present an improved wavelet-based embedded image coding scheme using vector quantization. The refinement process in wavelet-based embedded image coding is defined as a temporal domain, and dynamic index reordering is applied to the temporal domain of successive approximation vector quantization. In this method, the redundancy in consecutive angle or pattern transition is exploited. An efficient encoding and decoding algorithm is also presented to alleviate the computational complexity of the dynamic index reordering vector quantization. We investigate the index probability distribution of the compressed images using our proposed scheme and show PSNR performance improvement against the conventional successive approximation vector quantization.