Moving text region detection for broken text recovery
AssigneeSamsung Electronics Co., Ltd.
InventorsZhong Li, Yoon kyung Kim, Hunsop Hong, Seongnam Oh
Life of this patent
Estimating the term…
Abstract
One embodiment provides a method comprising receiving content for presentation on a display, and obtaining one or more sample frames of the content. The method further comprises generating a set of features based on one or more horizontal edge signals and one or more vertical edge signals of the one or more sample frames. The method further comprises utilizing a classification model to detect, based on the set of features, a region of interest of moving text in the one or more sample frames.
Claims20
A method comprising:
- receiving content for presentation on a display;
- obtaining one or more sample frames of the content;
- generating a set of features based on one or more horizontal edge signals and one or more vertical edge signals of the one or more sample frames;
- utilizing a classification model to classify the set of features with a classification identifying a content channel which the content originates from, wherein for each of a plurality of different content channels, the classification model is trained to learn different sets of features indicative of where and how moving text is displayed by the plurality of different content channels;
- and
- detecting, based on the classification, a region of interest of moving text in the one or more sample frames.
A system comprising:
- at least one processor;
- and
- a non-transitory processor-readable memory device storing instructions that when executed by the at least one processor causes the at least one processor to perform operations including:
- receiving content for presentation on a display;
- obtaining one or more sample frames of the content;
- generating a set of features based on one or more horizontal edge signals and one or more vertical edge signals of the one or more sample frames;
- utilizing a classification model to classify the set of features with a classification identifying a content channel which the content originates from, wherein for each of a plurality of different content channels, the classification model is trained to learn different sets of features indicative of where and how moving text is displayed by the plurality of different content channels;
- and
- detecting, based on the classification, a region of interest of moving text in the one or more sample frames.
A non-transitory computer readable medium that includes a program that when executed by a processor performs a method comprising:
- receiving content for presentation on a display;
- obtaining one or more sample frames of the content;
- generating a set of features based on one or more horizontal edge signals and one or more vertical edge signals of the one or more sample frames;
- utilizing a classification model to classify the set of features with a classification identifying a content channel which the content originates from, wherein for each of a plurality of different content channels, the classification model is trained to learn different sets of features indicative of where and how moving text is displayed by the plurality of different content channels;
- and
- detecting, based on the classification, a region of interest of moving text in the one or more sample frames.
Dependencies are inferred from the claim text.