An image processing apparatus includes: a reducing section reducing an image for which a feature analysis is to be perform at a predetermined reduction ratio; an ROI mask generating section analyzing a feature of a reduced image as the image reduced at the predetermined reduction ratio, and generating an ROI mask as mask information indicating a region of interest as a region to be interested in the reduced image; an ROI mask enlarging section enlarging a size of the ROI mask to a size of the image before being reduced by the reducing section; and an ROI mask updating section analyzing a feature of a region, set as a blank region as a region not to be interested in the ROI mask, of the image before being reduced by the reducing section, and updating the ROI mask by using an analysis result.