We propose a new context-based image coding technique using fuzzy logic in the quantization and entropy coding processes. Firstly, we have improved the well-known trellis-coded quantization (TCQ), using previous states of the TCQ symbols and fuzzy logic to predict the current TCQ state of the quantized wavelet coefficient. We designed the fuzzy rules and membership functions using our experiences in image coding. The quantized coefficients were presented with the bit planes, and the bits were coded arithmetically. Secondly, we propose a new probability estimation method for the adaptive arithmetic coder, using fuzzy logic to assign the probability of the coded symbol. Observations of previously coded bits were used as inputs of the fuzzy logic, which assigns the probability of the coded bit according to the fuzzy rules and membership functions. The efficiency of the context-based image coder using fuzzy logic achieves state-of-the-art results with respect to rate distortion.