Film frames, or other images in which fields are captured at the same point in time, may be processed as a sequence of temporally coherent image fields or as progressive images. Such images may be obtained, for example, by digitizing signals from a telecine and dropping redundant fields inserted by the telecine. These fields may be stored in a buffer. Two fields of a given frame are read from the buffer by a resizer in accordance with read instructions, which may be determined according to any specified pulldown sequence, an output image size, and resize instructions, such as pan and scan or letterbox instructions. The resizer also may be informed of the input image size if it is not presumed. Thus, the full input image from which an output image may be generated is used by the resizer to generate output image. The resizer uses data in the input image received at one rate to generate one output image at the output rate. The resize instructions, if varied over time, are determined for each output image at the output image rate. For example, if a pan and scan operation is specified to move from a first position to a second position over two input images, the position of the sampled area may be determined by interpolating between the first and second positions and sampling along the interpolated curve for each output image. Thus, the sampled area of the input image is different for each output field. Because changes in position of the sampled area are made in increments at the temporal resolution and ordering of the output images, artifacts in the output images are reduced.