Image matching method, extracting features from model images applied different variation, image matching device, model template generation method, and model template generation device
An image matching method performed by an image matching device, includes acquiring a model image, generating a plurality of variation images by applying to the model image variations that are different from each other, extracting features from each of the plurality of variation images, generating a model template based on the features, and performing template matching using an image to be matched and the model template.
1. An image matching method performed by an image matching device, comprising:
acquiring a model image;
generating a plurality of variation images by applying to the model image variations that are different from each other;
extracting features from each of the plurality of variation images;
generating a model template based on the features; and
performing template matching using an image to be matched and the model template,
wherein the extracting the features comprises segmenting the variation image into a plurality of regions and extracting the feature of each region, and
wherein the generating the model template comprises generating the model template using the features of the regions where a kurtosis of a feature distribution is higher than or equal to a predetermined threshold.
2. An image matching method performed by an image matching device, comprising:
acquiring a model image;
generating a plurality of variation images by applying to the model image variations that are different from each other;
extracting features from each of the plurality of variation images;
generating a model template based on the features; and
performing template matching using an image to be matched and the model template,
wherein the extracting the features comprises segmenting the variation image into a plurality of regions and extracting the feature of each region,
wherein the generating the model template comprises assigning an index value for the region to each of the regions based on a kurtosis of the distribution of the features extracted in the extracting the features, and
wherein the performing the template matching uses the index value assigned to each of the regions as a weight.
3. A model template generation method performed by a model template generation device, comprising:
acquiring a model image;
generating a plurality of variation images by applying to the model image variations that are different from each other;
extracting features from each of the plurality of variation images;
generating a model template based on the features; and
performing template matching using an image to be matched and the model template,
wherein the extracting the features comprises segmenting the variation image into a plurality of regions and extracting the feature of each region, and
wherein the generating the model template comprises generating the model template using the features of the regions where a kurtosis of a feature distribution is higher than or equal to a predetermined threshold.
4. A model template generation device comprising:
a model image acquisition unit that acquires a model image;
a variation image generator that generates a plurality of variation images by applying to the model image variations that are different from each other;
a feature extractor that extracts features from each of the plurality of variation images; and
a template generator that generates a model template based on the features,
wherein the extracting the features comprises segmenting the variation image into a plurality of regions and extracting the feature of each region, and
wherein the generating the model template comprises generating the model template using the features of the regions where a kurtosis of a feature distribution is higher than or equal to a predetermined threshold.