JP4896160B2 - Method for confirming conspicuous area of multimedia document, method for generating advertisement document, method for automatically confirming conspicuous area of Internet site, and related computer program - Google Patents

Description

The present invention relates to an apparatus and method for confirming a conspicuous area of a multimedia document.

  The invention applies in particular to documents such as content evaluation and advertising documents. Today, when you want to make content stand out, various techniques for generating content are known. To put it simply, when writing a text document, you may draw a line, if you write in bold There is a case where a frame is written. More sophisticated techniques are used by advertisers that can combine image and audio processing techniques, and more generally signal processing techniques, for video or audio.

  It is difficult to evaluate the effectiveness of these techniques. Advertisements are often quite attractive, but the various effects that are performed are not necessarily all effective and can be reliably improved.

Today, oculometric technology is useful in determining the conspicuous area. However, such a technique is manual and difficult to handle because it is not possible to automate techniques in multiple documents, web pages, and video documents.

To this end, the present invention provides a method for confirming the conspicuous area of a multimedia document. According to the present invention, the method comprises
-Determining an area of the document that is considered to be a key area based on prior data extracted from the document;
- so as to obtain a noticeable degree map of the document, and applying a model based on editing noticeable degree map in the document,
Comparing the region of the document that is considered to be a key with the conspicuity map;
Have

According to a preferred embodiment, the method comprises calculating a ratio of the conspicuity included in the key area of the document to the total conspicuity of the document.

According to a preferred embodiment, the method includes the steps of calculating the visual path indicating the direction of the eye during the viewing of the document, it said visual path is computed on the basis of the noticeable level map, the conspicuous degree Having pixels of the image that are greater than the threshold.

According to a preferred embodiment, during the search of the visual path,
-The pixel with the highest conspicuousness in the document is searched,
The area of the document around the maximum conspicuousness is invalidated;
The next pixel with the highest degree of visibility is searched until all pixels with a degree of visibility greater than the threshold are obtained.

Advantageously, an analysis report having a conspicuous region, the visual path and a set of sub-images representing the conspicuous region of the document are compiled.

  According to a preferred embodiment, the report has a key area for the document.

The invention also relates to a method for generating an advertising document. In accordance with the present invention, the method alternatively comprises the step of generating content followed by a method for confirming the salient area of the document according to the present invention, the step comprising the conspicuous area being a key area of the document. It is repeated unless it corresponds.

The present invention also relates to a method for automatically checking the conspicuous area of an Internet site. In accordance with the present invention, the method comprises the following steps for each page of the site:
-Determining an area of the page that is considered to be a key area of pre-data extracted from the page;
- so as to obtain a noticeable degree map of the page, and applying a model based on compiling noticeable degree map in the page,
-Comparing between the conspicuity map and the area of the page considered to be a key;
Have

  The invention also relates to a computer program product comprising program code instructions for execution of the steps of the method according to any one of the embodiments of the invention when said program is executed in a computer.

  The present invention is shown and fully understood by way of an illustrative embodiment and mode of implementation, which is entirely non-limiting, associated with FIG. 1, which represents a verification technique according to a preferred mode of the present invention. Can do.

  The present invention is particularly advantageous in documents with advertising properties where the components of the document are particularly important due to their economic impact. In particular, advertising films that are still images in video, simple sound or magazines are particularly sophisticated to produce messages that are attractive to consumers. However, sometimes these messages are not necessarily in the most attractive area of the human eye, in which case the advertiser has not created a document with the expected effect.

  The present invention therefore provides a means to analyze documents to ensure that they have a positive impact when they are received, whether or not they are advertising documents. With the goal.

Therefore, there is a method for calculating a conspicuousness map for a document. Those approaches make it possible to determine the eye-catching area. These approaches are based on operations that take into account various parameters of the image and on spatial and / or temporal analysis of the image and / or video.

The conspicuousness map is a topographic representation of image recognition. The visual relevance of each pixel of the image is determined. The map is normalized within a range of 0 to 255, for example. The larger the value, the more attractive the pixel is.

  Such a technique is, for example, by the applicant Thomson Licensing SA and is described in EP 1544792 published on June 22, 2005.

  Paper document D of advertising document type can be used. This document is created based on publishing software for which electronic versions of these software are available, for example, HTML type acrobat writer (registered trademark (by Thomson Licensing SA)), word (registered trademark (by Microsoft)) Has been. It is therefore possible to extract prior information from the document that reflects what the author of the document wanted to stand out. For example, in a word type document, it is possible to determine the importance of information as a function of the size of the cursor used, the font used, the font used, the line drawn over the character. Also, it is possible to use important information located in the sheet, generally in the center of the image.

  In html type documents, “tags” also assist in determining the points that the author wants to be prominent.

  Therefore, it is possible to compile the prior information representing the key points of document D by the author (step E1), either manually in the observation of document D or in the automatic method in the automatic extraction of features described above.

Therefore, it is possible to obtain the geometric position of the target area of the document.

Furthermore, it is possible to use a model that makes it possible to build a document conspicuousness map. This model is applied to document D during step E2. This model is, for example, software that accepts the document as input. The model analyzes the document features necessary for extracting the conspicuousness map so as to obtain the conspicuousness map of the document D.

  These techniques are described in, for example, European Patent No. 1544792 published on June 22, 2005 by Thomson Licensing SA. These approaches are based on operations that take into account various parameters of the image, and are based on spatial and / or temporal analysis of the image and / or video.

  Text type information and face type information are also typically extracted because they also represent information that is considered important when text is included in a video document.

For consideration of face-type or text-type information during compiling of the conspicuity map, refer to the document entitled “Attendion model based progress image transmission” by Hu, Xie, Chen, Ma, and published by Chen et al. It is described in the 2003 journal Journal “Multimedia Systems” entitled “A visual attentation model for adapting images on small displays”.

During step E3, based on the conspicuousness map, the following information:
-A heat map showing the attractive areas,
-A map of the visual path showing the visual direction of the average observer,
An analysis report is obtained.

This visual direction spans the conspicuity map for local maxima with a conspicuity value greater than a certain threshold as a function of the number of maxima desired to be obtained or as a function of the overall conspicuity of the image. Obtained by searching. For example, the search for local maxima is done by invalidating the circular area around the (n−1) th maxima (representing a fovea with a radius equal to diopter 0.5) and this new This is done by searching the global maximum, such as a map, and a search is made for each local maximum. For each maxima, therefore, regions are obtained, after which those regions n-1, n,. . . Are centered so as to obtain the average observer's visual direction.
A perceptual point with n sub-images of the image to be analyzed. These sub-images are areas fixed by the observer.

  A technique for determining such a sub-image is described in European Patent No. 05291938.8, filed Sep. 19, 2005, by the applicant Thomson Licensing SA. This method is also described in O.D. Le Meur, X. Castellan, P.A. Le Callet and D.C. It is described in the document ICIP 2006 entitled “Efficient saliency-based repurposing method” by Barba.

The sub-image has a local maximum surrounding pixel. The number of sub-images is determined by fixing a threshold value, and sub-images are defined only for the maximum having a conspicuousness value greater than this threshold value.

  This information makes it possible to detect regions of the image that are actually perceived by the human eye.

  Thereafter, the region indicated by the author of the document or automatically extracted during step E1 is compared with the region indicated in the analysis report. Therefore, by positioning the prior information and the information of the analysis report so as to face each other, it is possible to compare information that is actually captured by the human eye with what the author wants to stand out in the document.

The analysis report can also provide this analysis directly by performing a direct comparison step between the conspicuity map and the prior data. In particular, it is possible to compare the geometric position of the conspicuousness region and the region that the author considers important.

For example, it is possible to calculate the ratio between the degree of conspicuousness included in the area defined by the author of the document and the overall degree of conspicuousness . This ratio can be a measure of the relevance of document design. It is therefore possible to evaluate the value of the document with regard to the creator's eye-catching.

  The present invention can therefore advantageously provide a way to evaluate documents for advertising or advertising.

  The present invention can also be understood as a remote evaluation method for advertising documents. In practice, it is actually possible to consider a model that is available at clients and Internet sites that submit documents in electronic form so as to confirm the location / area that the human eye actually captures. The user of the service, as a result of the comparison, resubmits iteratively so that the information that the user wants to be prominent is exactly the information that the human eye focuses.

  The method can also have an evaluation of multiple documents submitted by the client as an application following a request for submission. Therefore, it is possible to automatically and objectively analyze which document has the most influence on the consumer.

  In other embodiments, document D is a video document, such as a promotion, sent to a television channel, rather than an ad insertion type paper document or still image. Such documents can also be analyzed according to the method implemented in the present invention.

In practice, it is also possible to compile a conspicuity map based on temporal information. For such methods, see O.D. Le Meur, P.A. Le Callet, D.C. Barba and D.W. Document Proc., Entitled “A spatial-temporal model of the selective human visual attention” by Thoreau. ICIP-05 (IEEE International Conference on Image Processing), Genoa, Italy, September 2005.

Therefore, it is also possible to obtain a noticeability map of the video document.

  It is possible to obtain information as a video author so as to obtain prior information regarding video data.

In other embodiments, it is also possible to use this approach by incorporating it directly into a tool that supports document creation. In that case, a document is actually obtained in which the eye focuses on a region that is actually considered a key in the document. Therefore, it is possible to create content as an alternative, apply the confirmation method according to the present invention, and modify the content as long as the conspicuous region and the key region do not match.

  In other embodiments, it is also possible to envisage documents that are parsed as being Internet pages that are parsed periodically, and each page of that site, either automatically or manually. Therefore, on all the web pages of the site, the method is carried out according to the present invention and the report, i.e. the data that the author considers to be key on the page, and if necessary or if key data is not available. It is possible to envisage an automated technique for blindly displaying the result of comparison with the data of the page that the eye actually sees.

Fig. 4 shows the steps of the method according to the preferred mode of the present invention.

Claims (9)

A method for checking the conspicuous area of a multimedia document:
Extracting an area of the document that is considered to be a key area based on pre-data extracted from the document , wherein the pre-data is made prominent by an author of the multimedia document Reflecting, steps ;
Applying to the document a model based on editing of a conspicuousness map to obtain a conspicuous region of the document; and automatically determining the geographical location of the region of the document considered to be a key and the conspicuousness map. the step of comparing the;
A method characterized by comprising: The method according to claim 1, further comprising a step of calculating a ratio of the degree of conspicuity included in the key area of the document to the overall degree of conspicuity of the document. 3. The method according to claim 1 or 2, wherein a visual path indicating an eye direction during visual recognition of the document is calculated, wherein the visual path is calculated based on the conspicuousness map, A method comprising the steps of: having an image pixel having a degree of conspicuity greater than a threshold value. The method of claim 3, wherein during the search for the visual path,
The pixel having the highest conspicuousness in the document is searched,
The area of the document around the maximum conspicuousness is invalid;
The next pixel having the maximum degree of conspicuousness is searched until all pixels having a degree of conspicuous greater than the threshold are obtained;
A method characterized by that. 5. A method as claimed in any preceding claim, wherein an analysis report comprising a set of sub-images representing the conspicuous region, the visual path and the conspicuous region of the document is compiled. And how to.   6. The method of claim 5, wherein the analysis report comprises the key area of the document. A method for creating an advertising document, the method comprising, as an alternative, a step of generating content, followed by the method for confirming the conspicuous area of the document according to any one of claims 1 to 6. The method of claim 1, wherein each step of claim 1 is repeated unless the conspicuous area corresponds to the key area of the document. A method for automatically checking the prominent area of an Internet site for each page of the site:
Determining the region of the page that is considered to be a key region based on pre-extracted data from the page, the pre-data making the author of the multimedia document prominent Reflecting, steps ;
So as to obtain a noticeable degree map of the page, a model based on editing noticeable degree map step is applied to the page; automatically comparing the area and the noticeable degree map of the page that are considered and key Step;
A method characterized by comprising: 9. A computer program comprising program code instructions for performing the steps of the method according to claim 1 when the program is executed on a computer.

Images