Image compression techniques based on subband coding are studied in this work. The analysis/synthesis algorithm is implemented using first-order all-pass recursive filters, which significantly reduces the computational complexity and reconstructs the input with neither aliasing nor phase distortion. Specific processing techniques for these filters were discussed. Limitations in direct subband coding show the convenience of initially splitting the image to be compressed into its low-pass and high-pass components, representing sharp edges, is divided into subbands which are vector quantized. Further improvement of this technique results from the study of subband characteristics and correlacion. Objective quality of each technique is measured by the peak signal-to-noise ratio and subjective quality results from visual inspection of reconstructed images. Both are superior or comparable to existing coders of similar complexity, for rates between 0.6 and 0.7 bits/pixel.