An image processing apparatus for encoding an image which contains a region to be encoded to have higher image quality than the surrounding portions, has detection means (100) for detecting a quantization step from a bitstream of the image encoded in a first stage, quantization step control means (103) for controlling the detected quantization step, and selection means (102) for encoding the image in a second stage using the controlled quantization step, and selecting one of the images encoded in the first and second stages.