The present invention relates to an image processing apparatus, an image processing method, and a program capable of obtaining, even if a process such as a graph cut method is performed in units of small regions, a result similar to that obtained in the process performed in units of pixels.A pre-segmentation executing unit 13 classifies pixels in an image on the basis of correlation among the pixels, thereby dividing the image into a set of small regions. A small region adjacent energy calculating unit 14 calculates a small region adjacent energy on the basis of a difference in pixel value of a pixel set at the vicinity of a boundary of a pair of small regions adjacent to each other in the set of small regions. A small region likelihood energy calculating unit 15 selects object regions and background regions for the respective small regions on the basis of a trimap image, constructs a probability model function, and calculates object likelihoods and background likelihoods for the respective small regions. A segmentation executing unit 16 divides the image into a binary mask image of a foreground and a background using a small region adjacent energy and a small region likelihood energy. The present invention can be applied to image processing of separating an object image.