Subject region detecting apparatus

An image processing method comprising:

• identifying a plurality of candidates of lines that constitute a contour of a subject region, on the basis of a captured image including an image of a subject captured with an imaging apparatus, the subject region corresponding to the image of the subject;
• calculating, for each of the plurality of identified candidates, line probability information representing a probability that the candidates are a line that constitutes the contour of the subject region, on the basis of feature values that are determined on the basis of the captured image and that are of positions located at or within a reference distance from the candidate;
• calculating region probability information representing a probability that a target region surrounded by a predetermined number of candidates chosen from the plurality of identified candidates is the subject region, on the basis of pieces of line probability information of the predetermined number of candidates;
• and detecting the subject region on the basis of the calculated region probability information,
• wherein the calculating region-probability information comprises obtaining, for one side of two adjacent sides that constitute a target region, feature values that are determined on the basis of the captured image and that are of positions located at or within a threshold distance from the side, and calculating the region probability information of the target region on the basis of the pieces of line probability information of the predetermined number of candidates and the obtained feature values, and
• wherein the calculating region-probability information further comprises generating information regarding similarity between the feature values obtained for one of the two adjacent sides and the feature values obtained for the other of the two adjacent sides, and calculating the region probability information on the basis of the pieces of the line probability information of the predetermined number of candidates and the generated information.

A non-transitory computer readable medium storing a program causing a computer to execute a process for image processing, the process comprising:

• identifying a plurality of candidates of lines that constitute a contour of a subject region, on the basis of a captured image including an image of a subject captured with an imaging apparatus, the subject region corresponding to the image of the subject;
• calculating, for each of the plurality of identified candidates, line probability information representing a probability that the candidates are a line that constitutes the contour of the subject region, on the basis of feature values that are determined on the basis of the captured image and that are of positions located at or within a reference distance from the candidate;
• calculating region probability information representing a probability that a target region surrounded by a predetermined number of candidates chosen from the plurality of identified candidates is the subject region, on the basis of pieces of line probability information of the predetermined number of candidates;
• and detecting the subject region on the basis of the calculated region probability information,
• wherein the calculating region-probability information comprises obtaining, for one side of two adjacent sides that constitute a target region, feature values that are determined on the basis of the captured image and that are of positions located at or within a threshold distance from the side, and calculating the region probability information of the target region on the basis of the pieces of line probability information of the predetermined number of candidates and the obtained feature values, and
• wherein the calculating region-probability information further comprises generating information regarding similarity between the feature values obtained for one of the two adjacent sides and the feature values obtained for the other of the two adjacent sides, and calculating the region probability information on the basis of the pieces of the line probability information of the predetermined number of candidates and the generated information.