To accurately identify a belonging body tissue of each portion in a body tissue image, based on a feature of a texture of the body tissue.
An image of an object to be identified is divided into a local area, in a step S11; a pixel block of 3×3 pixels around the respective pixel of the local area as the center is extracted, in a step S12. The pixels in each pixel block are encoded into an 8-bit code, in a step S13, and a local area histogram corresponding to the local area is generated to be normalized, in a step S14. A Euclidean distance is computed between a feature histogram (58-dimensional vector) of each body tissue generated preliminarily to be held, and the local area histogram (58-dimensional vector) of each local area in the identification-objective image, in a step S15, and the body tissue in each portion area of the identification-objective image, based on whether the distance is a prescribed threshold value or less, or not. The image processor is applicable, for example, to an image analyzer in a medical field.
Takashi Nishikawa
Next Patent: MIST DEVICE