Wireless capsule endoscope is a state-of-the-art tool for detecting intestines problems. The amount of image data generated by using capsule endoscope is so large that data compression is desirable. In this work, we propose a fast and accurate quality-controlling algorithm that does not require the recursive estimation of distortion threshold for the compression of capsule endoscope images. A corresponding distortion in wavelet domain can be estimated directly from a user-defined PSNR (peak signal to noise ratio) in pixel domain, resulting in the distortion threshold that is necessary for the codebook replenishment mechanism in a wavelet-based adaptive vector quantizer (AVQ). Furthermore, in our AVQ implementation, a modeling rather than a training technique is proposed to generate an initial codebook (CB), where the parameters of the model are produced by using a pseudo-noise sequence in order to create such a CB at both the encoder and the decoder. To enhance the error resilient capability of AVQ in an error-prone channel, an efficient compression scheme based on novel tree blocks in wavelet domain is proposed, and look-up-table (LUT) and codebook replenishment (CBR) mechanisms in AVQ are replaced by motion estimation (ME) and motion compensation (MC) mechanisms in the proposed method, respectively. Experimental results show that the proposed method does give a fast and reliable quality control of all reconstructed capsule endoscope images under test, and the CB modeling produces comparable performance to the one using CB training. The proposed coding scheme and its AVQ predecessor have comparable coding performance but the former is more robust to transmission errors and more efficient in memory requirement.