Adaptive image processing apparatus and method based in image pyramid
View Patent ↗An adaptive image processing apparatus includes a first matching unit selecting representative candidate images from among a plurality of images included in an image pyramid and calculating a first matching score between each of the representative candidate images and a target model; a second matching unit selecting one scale level from among scale levels of the representative candidate images based on the first matching score and calculating a second matching score between each of images having scale levels included in a dense scale level range with respect to the selected scale level and the target model; a representative scale level selecting unit selecting at least one of the scale levels included in the dense scale level range as a representative scale level based on the second matching scores; and an image processing unit performing image processing based on an image having the selected representative scale level.
1. An adaptive image processing apparatus comprising:
a first matching unit which selects a plurality of representative candidate images from among a plurality of images included in a image pyramid so that scale levels of the selected plurality of representative candidate images are differentiated over a coarse interval and calculates a first matching score between each of the plurality of representative candidate images and a target model;
a second matching unit which selects one scale level from among the scale levels of the plurality of representative candidate images based on the first matching score and calculates a second matching score between the target model and each of images that have the selected scale level and other scale levels differentiated over a fine interval from the selected scale level in a dense scale level range, the fine interval being narrower than the coarse interval;
a representative scale level selecting unit which selects at least one of the scale levels included in the dense scale level range as a representative scale level based on the second matching scores; and
an image processing unit which performs image processing based on an image having the selected representative scale level.
2. The apparatus of claim 1 , wherein the first matching unit selects the plurality of representative candidate images to be separated apart by the same coarse interval in an alignment order in the image pyramid.
3. The apparatus of claim 2 , wherein the dense scale level range is determined to include a predefined number of scale levels above and below the scale level selected by the second matching unit in the alignment order in the image pyramid.
4. The apparatus of claim 1 , wherein the scale levels included in the dense scale level range are determined based on at least one of a number of the plurality of images having different scale levels in the image pyramid, a number of the plurality of representative candidate images, and the coarse interval between the plurality of representative candidate images in the alignment order in the image pyramid.
5. The apparatus of claim 1 , wherein the first matching unit extracts representative candidate images as samples by using a systematic sampling method.
6. The apparatus of claim 1 , wherein the second matching unit selects the scale level of a representative candidate image having a highest first matching score from among the scale levels of the plurality of representative candidate images.
7. The apparatus of claim 1 , wherein the representative scale level selecting unit selects, as a representative scale level:
a scale level having a second matching score equal to or greater than a threshold value set according to an algorithm for the image processing from among the scale levels included in the dense scale level range; or
a scale level having a highest second matching score from among the scale levels included in the dense scale level range.
8. The apparatus of claim 1 , wherein the first matching score is calculated by matching each of the plurality of representative candidate images to the target model by using a convolution algorithm or a correlation coefficient-based algorithm.
9. The apparatus of claim 1 , wherein the image processing unit extracts an object related to the target model from an image of the representative scale level.
10. The apparatus of claim 1 , wherein the target model is a model formed by statistically learning feature information of at least one object from among a human, a non-human animal, and a vehicle.
11. An adaptive image processing method comprising:
selecting a plurality of representative candidate images from among a plurality of images included in an image pyramid and having different scale levels so that the difference scale levels of the selected plurality of representative candidate images are differentiated over a coarse interval;
calculating a first matching score between each of the plurality of representative candidate images and a target model;
selecting one scale level from among the scale levels of the plurality of representative candidate images based on the first matching score;
calculating a second matching score between the target model and each of images that have the selected scale level and other scale levels differentiated over a interval from the selected scale level in a dense scale level range, the fine interval being narrower than the coarse interval;
selecting at least one from among the scale levels included in the dense scale level range based on the second matching scores as a representative scale level; and
performing image processing based on an image having the representative scale level.
12. The method of claim 11 , wherein the plurality of representative candidate images are selected to be separated apart by the same coarse interval in an alignment order in the image pyramid.
13. The method of claim 12 , wherein the dense scale level range is determined to include a predefined number of scale levels above and below the scale level selected in the selecting one scale level from among scale levels of the plurality of representative candidate images based on the first matching score, in the alignment order in the image pyramid.
14. The method of claim 11 , wherein the scale levels included in the dense scale level range are determined based on at least one of a number of the plurality of images having different scale levels in the image pyramid, a number of the plurality of representative candidate images, and the coarse interval between the plurality of representative candidate images in the alignment order in the image pyramid.
15. The method of claim 11 , wherein the plurality of representative candidate images are samples extracted by using a systematic sampling method.
16. The method of claim 11 , wherein the scale level, selected in the selecting one scale level from among scale levels of the plurality of representative candidate images based on the first matching score, is a scale level of a representative candidate image having a first highest matching score from among the scale levels of the plurality of representative candidate images.
17. The method of claim 11 , wherein the representative scale level is:
a scale level having a second matching score equal to or greater than a threshold value set according to an algorithm for the image processing from among the scale levels included in the dense scale level range; or
a scale level having a highest second matching score from among the scale levels included in the dense scale level range.
18. The method of claim 11 , wherein the first matching score is calculated by matching each of the plurality of representative candidate images to the target model by using a convolution algorithm or a correlation coefficient-based algorithm.
19. The method of claim 11 , wherein the performing the image processing comprises extracting an object related to the target model from an image of the representative scale level.
20. The method of claim 11 , wherein the target model is a model formed by statistically learning feature information of at least one object from among a human, a non-human animal, and a vehicle.