We present objective and subjective evaluations of five adaptive speckle reduction techniques: median filter, Lee multiplicative filter, geometrical filter, anisotropic diffusion and polynomial transform-based methods. For the objective evaluations we used several local and global measures of contrast and signal-to-noise ratio. For the subjective evaluation we used numerical scaling experiments assessing basic perceptual attributes. These are used to construct a multidimensional perceptual space subtended by perceptual attributes, namely noisiness, sharpness and contrast. These attributes, together with quality, are represented as vectors in the space. Results of the subjective evaluation show that quality of the processed images is mainly determined by the perception of noise, which in turn influences the perception of sharpness and contrast. The geometrical filter appears as the best filter in terms of objective measures, while the polynomial transform method is the best filter according to the subjective evaluation.