Patent US 6,567,558

An image decoder includes a demultiplexer which extracts motion vector information and quantized DCT coefficients from input information to be decoded, a dequantizer which dequantizes the quantized DCT coefficients to obtain DCT coefficients, an inverse DCT converter which performs inverse DCT conversion on the DCT coefficients to obtain an error image, a prediction image synthesizer which synthesizes a prediction image by performing motion compensation using the motion vector information and a reference image which is a previously decoded image, and an adder which adds the prediction image to the error image to obtain a decoded image. The motion compensation is performable using a positive rounding method and a negative rounding method for interpolating intensity values of pixels in performing the motion compensation.