Vector quantization (VQ) is an efficient technique for signal compression. However, it requires much encoding time to find the closest codeword for every input vector. We propose a fast encoding method to speed up the encoding. With the help of a table that is created off-line and can be used by all the images, the encoder searches only part of the entire codebook. The proposed method is implemented to encode Lena and other images to test its performance. Compared to full-searching VQ (FS-VQ), although the encoder searches only about 20 codewords in the codebook for every input vector, more than 95% of the codewords searched by the proposed method are the same as the results searched by FS-VQ on average. In addition, we also adopt partial distortion searching (PDS) and lookup table (LUT) to decrease the mathematic computation. This saves 98.44% of the encoding time and 98.07% of the mathematic operation while encoding Lena. The proposed method is superior to all the existing fast VQ encoding methods. While encoding 100 nature images for testing, it can save more than 97% of the encoding time and mathematic operations, but the PSNR decays at most only 0.19 dB, which is invisible to human eyes.