US 2022/0092338 A1Published application

Image labeling apparatus, method, and non-transitory computer readable storage medium thereof

AssigneeInstitute for Information Industry

InventorsChing-Han YANG, Deron LIANG, Chin-Chun CHANG

Life of this patent

Estimating the term…

Loading patent…

Abstract

An image labeling apparatus, method, and computer program product are provided. The image labeling apparatus generates several image features of several to-be-processed images and classifies them into several groups. For each group, the image labeling apparatus: (a) selects the image feature closest to the center as a representative feature and a compared feature, (b) selects the image feature farthest to the compared feature as a candidate feature, and (c) compares the similarity between the candidate feature and each representative feature with a threshold. When all the similarities are smaller than the threshold, adopts the candidate feature as another representative feature and the compared feature and repeats (b) and (c). The group(s) whose number of representative features is more than another threshold will be re-classified. The image labeling apparatus determines several to-be-labeled groups according to the classification result and labels the to-be-processed images within the same to-be-labeled group the same.

Claims20

Dependent claims

An image labeling apparatus, comprising:

a transceiving interface;
a storage, being configured to store a plurality of to-be-processed images;
and
a processor, being electrically connected to the transceiving interface and the storage and configured to generate an image feature for each of the to-be-processed images, classify the image features into a plurality of groups, and perform the following operations on each of the groups to generate at least one representative feature of each of the groups:
(a) selecting the image feature that is closest to a center of the group from the image features of the group as one of the at least one representative feature of the group and as a first compared feature,
(b) selecting the image feature that is farthest to the first compared feature from the image features of the group that have not been selected as a first candidate feature, wherein a first similarity degree is between the first candidate feature and each representative feature of the group, and
(c) comparing at least one of the at least one first similarity degree with a first threshold,
wherein when the at least one first similarity degree is all smaller than the first threshold, the processor assigns the first candidate feature as one of the at least one representative feature of the group and as the first compared feature of the next iteration and performs the operation (b) and the operation (c) again,
wherein the groups are classified into a first portion and a second portion by a second threshold, a first number of the at least one representative feature of each group of the second portion is greater than the second threshold, the processor classifies each group of the second portion into a plurality of refined groups and assigns the groups of the first portion and the refined groups of the second portion as a plurality of to-be-labeled groups,
wherein the transceiving interface receives a plurality of labeling instructions, each labeling instruction corresponds to one of the to-be-labeled groups, and the processor labels the to-be-processed images of each to-be-labeled group according to the corresponding labeling instruction.

7 more elements

An image labeling method for use in an electronic computing apparatus, the electronic computing apparatus storing a plurality of to-be-processed images, the image labeling method comprising:

(a) generating an image feature for each of the to-be-processed images;
(b) classifying the image features into a plurality of groups;
(c) performing the following operations on each of the groups to generate at least one representative feature of each of the groups:
(c1) selecting the image feature that is closest to a center of the group from the image features of the group as one of the at least one representative feature of the group and as a first compared feature;
(c2) selecting the image feature that is farthest to the first compared feature from the image features of the group that have not been selected as a first candidate feature, wherein a first similarity degree is between the first candidate feature and each representative feature of the group;
and
(c3) comparing at least one of the at least one first similarity degree with a first threshold,
wherein when the at least one first similarity degree is all smaller than the first threshold, assigning the first candidate feature as one of the at least one representative feature of the group and as the first compared feature of the next iteration and performing the step (c2) and the step (c3) again,
wherein the groups are classified into a first portion and a second portion by a second threshold, a first number of the at least one representative feature of each group of the second portion is greater than the second threshold, and the image labeling method further comprises the following steps of:
(d) classifying each group of the second portion into a plurality of refined groups;
(e) assigning the groups of the first portion and the refined groups of the second portion as a plurality of to-be-labeled groups;
(f) receiving a plurality of labeling instructions, wherein each labeling instruction corresponds to one of the to-be-labeled groups;
and
(g) labeling the to-be-processed images of each to-be-labeled group according to the corresponding labeling instruction.

11 more elements

A non-transitory computer readable storage medium, storing a computer program comprising a plurality of codes, the computer program executing an image labeling method after the codes are loaded into an electronic computing apparatus, the electronic computing apparatus storing a plurality of to-be-processed images, and the image labeling method comprising:

(a) generating an image feature for each of the to-be-processed images;
(b) classifying the image features into a plurality of groups;
(c) performing the following operations on each of the groups to generate at least one representative feature of each of the groups:
(c1) selecting the image feature that is closest to a center of the group from the image features of the group as one of the at least one representative feature of the group and as a first compared feature;
(c2) selecting the image feature that is farthest to the first compared feature from the image features of the group that have not been selected as a first candidate feature, wherein a first similarity degree is between the first candidate feature and each representative feature of the group;
and
(c3) comparing at least one of the at least one first similarity degree with a first threshold,
wherein when the at least one first similarity degree is all smaller than the first threshold, assigning the first candidate feature as one of the at least one representative feature of the group and as the first compared feature of the next iteration and performing the step (c2) and the step (c3) again,
wherein the groups are classified into a first portion and a second portion by a second threshold, a first number of the at least one representative feature of each group of the second portion is greater than the second threshold, and the image labeling method further comprises the following steps of:
(d) classifying each group of the second portion into a plurality of refined groups;
(e) assigning the groups of the first portion and the refined groups of the second portion as a plurality of to-be-labeled groups;
(f) receiving a plurality of labeling instructions, wherein each labeling instruction corresponds to one of the to-be-labeled groups;
and
(g) labeling the to-be-processed images of each to-be-labeled group according to the corresponding labeling instruction.

11 more elements

Dependencies are inferred from the claim text.

Facts

Application: 17/078,089Filed 22 Oct 2020
Priority: TW 10913259921 Sep 2020
Published: 24 Mar 2022
Term ends: Not estimatedThe source records don’t support an estimate.

Classification11 codes

G06F 18/2431
Relating to the number of classes: multiple classes
G06V 10/82
Using pattern recognition or machine learning: using neural networks
G06F 18/211
Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation: selection of the most…
G06F 18/22
Analysing: matching criteria, e.g. proximity measures