Image processing apparatus, method, and program

An image processing method of an image processing apparatus including:

• a feature vector extraction unit that extracts feature vectors of sample pixels, for each local region including the sample pixels, in an input image;
• a weight calculation unit that calculates weights for the local regions on the basis of the positional relationship with the local regions, for divided regions that divide the input image;
• a feature distribution model calculation unit that calculates an average with weights as a feature distribution model in the local region obtained by adding the weights calculated for the local regions to the feature vectors of the local regions, for each of the divided regions;
• and a pixel likelihood calculation unit that calculates pixel likelihood for each pixel in the input image on the basis of the feature distribution model,
• the method comprising:
• extracting feature vectors of the sample pixels, for each of the local regions including the sample pixels in the input image, with the feature vector extraction unit;
• calculating weight for each of the local regions on the basis of the positional relationship with the local regions, for each of the divided regions of the input image, with the weight calculation unit;
• calculating a average with weights obtained by adding the weights calculated for each of the local regions to the feature vectors of the local regions, for each of the divided regions, as a feature distribution models of the local regions, with the feature distribution model calculation unit;
• and calculating pixel likelihood for each of the pixels in the input image on the basis of the feature distribution model, with the pixel likelihood calculation unit.

A program performing a process including:

• extracting feature vectors of the sample pixels, for each of the local regions including the sample pixels in the input image, with the feature vector extraction unit;
• calculating weight for each of the local regions on the basis of the positional relationship with the local regions, for each of the divided regions of the input image, with the weight calculation unit;
• calculating a average with weights obtained by adding the weights calculated for each of the local regions to the feature vectors of the local regions, for each of the divided regions, as a feature distribution model of the local regions, with the feature distribution model calculation unit;
• and
• calculating pixel likelihood for each of the pixels in the input image on the basis of the feature distribution model, with the pixel likelihood calculation unit,
• in a computer controlling an image processing apparatus including:
• a feature vector extraction unit that extracts feature vectors of sample pixels, for each local region including the sample pixels, in an input image;
• a weight calculation unit that calculates weights for the local regions on the basis of the positional relationship with the local regions, for divided regions that divide the input image;
• a feature distribution model calculation unit that calculates an average with weights as a feature distribution model in the local region obtained by adding the weights calculated for the local regions to the feature vectors of the local regions, for each of the divided regions;
• and a pixel likelihood calculation unit that calculates pixel likelihood for each pixel in the input image on the basis of the feature distribution model.