Image analysis – Applications
Reexamination Certificate
2000-01-10
2003-12-16
Boudreau, Leo (Department: 2621)
Image analysis
Applications
Reexamination Certificate
active
06665419
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a digital watermark inserting system for inserting digital watermark information into an input image, a digital watermark characteristic parameter table generating method, and a computer readable record medium on which a digital watermark characteristic parameter table generating program has been recorded.
2. Description of the Related Art
In recent years, digital watermarks are becoming attractive. However, there have been few studies that focus on attacks against digital watermarks. As a result, it is difficult to compare robustness values against attacks in different digital watermark systems. As a related art reference, a framework of a robustness evaluation value calculating system for categorizing attacks that take place in conventional image processing and so forth and calculating robustness evaluation values for individually categorized attacks has been disclosed by Ryoma Oami, Yoshihiro Miyamoto, and Mutsumi Ohta, NEC C & C Media Research Laboratories in “Robustness Measure against attacks for digital watermarking and its application,” 1998 Image Media Processing Symposium (IMPS 98). The related art reference describes a method for obtaining the optimum digital watermark strength.
In the related art reference, attacks against digital watermarks are largely categorized as (1) deterioration that takes place in an image processing or the like and (2) intentional forgery of embedded information. Since attacks of the category (1) inevitably take place in a conventional image process. Thus, strong robustness is required for attacks of the category (1). Attacks of the category (2) are higher level attacks than those of the category (1). The attacks of the category (2) largely depend on the digital watermark inserting and detecting system. In the related art reference, attacks of the category (1) are further categorized from a view point of a signal processing. Robustness evaluation values are calculated for individually categorized attacks. The categorized attacks are for example coding loss in JPEG (LC), uniform and Gauss type noise (N), geometric transform such as scale and rotation (GT), pixel value conversion such as gray and binary (PVC), and image processing such as sharpening and median filtering (IP).
In the system of the related art reference, a digital watermark is inserted into an image. Thus, a digital watermark inserted image is obtained. Thereafter, the digital watermark inserted image is attacked in a predetermined manner and then the digital watermark is detected. By repeating the above procedures, varying a parameter for adjusting the attack strength, a detection ratio is obtained. The detection ratio data is statistically processed. Thus, robustness evaluation values for individually categorized attacks are calculated. As the statistical process, weighted means method or threshold value method is used, for example. In the evaluation value calculating system, (a) by properly setting a weighting function, evaluation values for applications having different attack characteristics can be calculated; (b) robustness evaluation values can be compared among different systems regardless of the values of the attack strength x for really measuring detection ratios; and (c) the accuracy of evaluation values can be improved progressively. According to the evaluation values calculating system, evaluation values 0 to 1 against watermark strength values 1 to 4 depend on conventional digital watermark systems.
However, according to the digital watermark inserting system of the related art reference, it is difficult to properly set the strength of a digital watermark. The strength of the digital watermark largely depends on the contents of the image. Even if the user designates the relation of the strength of digital watermarks, the deterioration of the image quality, and the robustness values against attacks in a deterministic manner, the relation cannot be applied to all images. Thus, the optimum strength of digital watermarks cannot be obtained.
The inventor of the present invention has invented “digital watermark inserting system” that is currently being filed as Japanese Patent Application No. 10-150823 (hereinafter, this invention may be referred to as second related art reference). This invention was made to embed (insert) copyright information and so forth into digital signals of audio data, image data, and so forth.
FIG. 1
is a block diagram showing the structure of the “digital watermark inserting system” as the second related art reference. In
FIG. 1
, a categorizing portion
103
calculates a feature amount of an input image, obtains a category of the image with the calculated feature amount, and outputs the index indicating the category as a category index to a storing unit
2001
. The storing unit
2001
selects a table corresponding to the category index that is received from the categorizing portion
103
and outputs an image quality deterioration ratio and a robustness evaluation value corresponding to digital watermark strength that is received from a digital watermark strength calculating portion
100
to the digital watermark strength calculating portion
100
.
The digital watermark strength calculating portion
100
outputs various digital watermark strength values to the storing unit
2001
. The digital watermark strength calculating portion
100
decides the optimum digital watermark strength based on the image quality deterioration ratio and the robustness evaluation value that are received from the storing unit
2001
and based on restriction information of digital watermark strength that is input by the user and outputs the decided optimum digital watermark strength data to a digital watermark inserting portion
102
.
The digital watermark inserting portion
102
converts embedding data into digital watermark information, inserts the digital watermark into the image with the optimum digital watermark strength received from the digital watermark strength calculating portion
100
, and outputs a digital watermark inserted image.
Next, the operation of the digital watermark inserting system shown in
FIG. 1
will be described. First of all, several symbols used in the operation will be defined.
The number of categories of input images is denoted by K. K categories are distinguished with category index k (where k=1, . . . , K). The digital watermark strength with which a digital watermark is inserted is denoted by s(m) (where m=1, . . . , M). The parameter used as the digital watermark strength depends on the digital watermark inserting algorithm for use. When the digital watermark strength is successively varied, it is digitized into M different values and their values are denoted by s(m). When the category index is k and the digital watermark strength is s(m), the image quality deterioration ratio and the robustness evaluation value against the attack are denoted by D(k,m) and V(k,m), respectively.
Next, with reference to
FIG. 1
, the operation of the digital watermark inserting system will be described. An input image is supplied to the categorizing portion
103
. The categorizing portion
103
calculates a feature amount of the image, decides the category of the input image based on the obtained feature amount, and outputs a category index that represents the category. In reality, the categorizing portion
103
stores feature amount values that represent boundaries of categories. The categorizing portion
103
compares the stored feature amount values with the calculated feature amount value and categorizes the input image based on the compared result. The feature amount is for example an activity of the entire image (the activity is the mean value of AC frequency components).
The category index that is output from the categorizing portion
103
in input to the storing unit
2001
. The storing unit
2001
stores digital watermark feature tables for individual category indexes. Each of the digital watermark feature tables represents the relation o
Akhavannik Hussein
Boudreau Leo
NEC Corporation
Scully Scott Murphy & Presser
LandOfFree
Digital watermark inserting system and digital watermark... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Digital watermark inserting system and digital watermark..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Digital watermark inserting system and digital watermark... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3149899