The noise effects in color images are studied from the human perception and machine perception point of view. Three justifiable observations are made to illustrate problems related to individual color signal processing. To minimize the noise effects, two solutions are studied: One is a ‘rental scheme’ and the other is a vector signal processing technique. The ‘rental scheme’ adopts filters originally developed for grey scale images to color images. A set of heuristic criteria is defined to reconstruct an output with minimum artifacts. The vector signal processing technique utilizes a median vector filter based on the well developed median filter for grey scale images. Since the output of the filter does not have the same physical meaning as the median defined in one-dimensional space, the search of a vector median is considered as a minimum problem. The output is guaranteed to be one of the inputs. Both approaches are shown to be very effective in removing speckle noise. Results from real and synthetic images are obtained and compared.