Abstract: This paper has described a new image rotation algorithm which can be easily implemented by hardware. The algorithm is executed by decomposing the image rotation transfomation into three sucessive sub-trans-formtion(row—col—row),each of them called pseudo-translation.As the pixels on one row (or column) is translated by a same offset,the translation between different rows or column can be accessed parallelly in every pseudo-translation.So the rotating time is shorter than conventional techniques.With this method,the processing time that it takes to rotate a gray scale image made up of 256×256 pixels is about 5s even performing in software.