Detection Apparatus, Detection Method, and Computer Program

Abstract

An apparatus for detecting a cut change based on a similarity between a first image and a second image, includes a unit for generating one of a luminance histogram and a color histogram of each of the first image and the second image, a unit for generating a spatial correlation image representing a correlation between spatial layouts of the first image and the second image, a unit for calculating a histogram similarity representing a similarity between the histogram of the first image and the histogram of the second image, a unit for calculating a spatial correlation image similarity representing a similarity between the spatial correlation image of the first image and the spatial correlation image of the second image, and a unit for determining whether a border between the first image and the second image is a cut change based on the histogram similarity and the spatial correlation image similarity.

1 . A computer program for causing a computer to detect a cut change based on a similarity between a first image and a second image, comprising steps of:
 generating one of a luminance histogram and a color histogram of each of the first image and the second image;   generating a spatial correlation image representing a correlation between spatial layouts of the first image and the second image;   calculating a histogram similarity representing a similarity between the histogram of the first image and the histogram of the second image;   calculating a spatial correlation image similarity representing a similarity between the spatial correlation image of the first image and the spatial correlation image of the second image; and   determining whether a border between the first image and the second image is a cut change based on the histogram similarity and the spatial correlation image similarity.   
   
   
       2 . The computer program according to  claim 1 , wherein the histogram similarity is an overlapping rate between the histogram of the first image and the histogram of the second image. 
   
   
       3 . The computer program according to  claim 1 , wherein the spatial correlation image generating step comprises generating, as the spatial correlation image, at least one of a filter image and a mosaic image, the filter image being an image that is obtained by reducing a low frequency component and a high frequency component from each of the first image and the second image, and the mosaic image composed of a plurality of areas, the entire image of each of the first image and the second image being partitioned into the plurality of areas, and each area being represented by one of the average luminance of the area and the average color of the area. 
   
   
       4 . The computer program according to  claim 3 , wherein the spatial correlation image is the filter image, and
 wherein the spatial correlation image similarity is a maximum value of correlation values being calculated by shifting in whole or in part the spatial correlation images of the first image and the second image in relative position.   
   
   
       5 . The computer program according to  claim 3 , wherein the spatial correlation image is the filter image, and
 wherein the spatial correlation image similarity is one of a minimum value of a sum of absolute differences and a minimum value of a sum of squared differences being calculated by shifting in whole or in part the spatial correlation images of the first image and the second image in relative position.   
   
   
       6 . The computer program according to  claim 3 , wherein the spatial correlation image is the mosaic image, and
 wherein the spatial correlation image similarity is one of the number of corresponding pixels, between the spatial correlation images of the first image and the second image, having a difference equal to or lower than a predetermined threshold, and the ratio of the number of corresponding pixels having the difference equal to or lower than the predetermined threshold to all the pixels.   
   
   
       7 . The computer program according to  claim 1 , wherein the determining step comprises determining whether the border between the first image and the second image is a cut change by comparing with a predetermined threshold value a value resulting from weighted summing the histogram similarity and the spatial correlation image similarity. 
   
   
       8 . The computer program according to  claim 7 , wherein the determining step comprises determining whether the border between the first image and the second image is a cut change by comparing with a predetermined threshold a value resulting from weighted summing a non-linear histogram similarity and a non-linear spatial correlation image similarity, the histogram similarity being converted into the non-linear histogram similarity using a non-linear function and the spatial correlation image similarity being converted into the non-linear spatial correlation image similarity using a non-linear function. 
   
   
       9 . The computer program according to  claim 1 , further comprising a step of calculating an average similarity being at least one of an average of histogram similarities of images in a plurality of adjacent frames calculated in the histogram similarity calculating step and an average of spatial correlation image similarities of the images in a plurality of adjacent frames calculated in the spatial correlation image similarity calculating step,
 wherein the determining step includes determining whether the border between the first image and the second image is a cut change based on the histogram similarity, the spatial correlation image similarity, and the average similarity.   
   
   
       10 . The computer program according to  claim 1 , further comprising a step of calculating a proximity average color being an average of luminance or color of the images on a plurality of frames adjacent to the first image,
 wherein the determining step includes determining whether the border between the first image and the second image is a cut change based on the histogram similarity, the spatial correlation image similarity, and the proximity average color.   
   
   
       11 . The computer program according to  claim 1 , further comprising steps of:
 generating a fine histogram of one of luminance and color of each of the first image and the second image;   calculating a feature distribution by filtering the fine histogram;   calculating a similarity between the feature distribution of the first image and the feature distribution of the second image;   extracting a feature of a third image different from the first image and the second image;   calculating a similarity between two images from among the first image, the second image and the third image;   generating a contracted image of each of the first image, the second image and the third image;   generating a synthesized image of the contracted image of the first image and the contracted image of the third image; and   calculating a similarity between the synthesized image and the contracted image of the second image,   wherein the determining step includes determining whether the border between the first image and the second image is a cut change based on the similarity between the feature distributions, the similarity between the two images from among the first image, the second image and the third image, and the similarity between the synthesized image and the contracted image of the second image.   
   
   
       12 . A method of detecting a cut change based on a similarity between a first image and a second image, comprising steps of:
 generating one of a luminance histogram and a color histogram of each of the first image and the second image;   generating a spatial correlation image representing a correlation between spatial layouts of the first image and the second image;   calculating a histogram similarity representing a similarity between the histogram of the first image and the histogram of the second image;   calculating a spatial correlation image similarity representing a similarity between the spatial correlation image of the first image and the spatial correlation image of the second image; and   determining whether a border between the first image and the second image is a cut change based on the histogram similarity and the spatial correlation image similarity.   
   
   
       13 . An apparatus for detecting a cut change based on a similarity between a first image and a second image, comprising:
 means for generating one of a luminance histogram and a color histogram of each of the first image and the second image;   means for generating a spatial correlation image representing a correlation between spatial layouts of the first image and the second image;   means for calculating a histogram similarity representing a similarity between the histogram of the first image and the histogram of the second image;   means for calculating a spatial correlation image similarity representing a similarity between the spatial correlation image of the first image and the spatial correlation image of the second image; and   means for determining whether a border between the first image and the second image is a cut change based on the histogram similarity and the spatial correlation image similarity.   
   
   
       14 . An apparatus for detecting a cut change based on a similarity between a first image and a second image, comprising:
 a unit generating one of a luminance histogram and a color histogram of each of the first image and the second image;   a unit generating a spatial correlation image representing a correlation between spatial layouts of the first image and the second image;   a unit calculating a histogram similarity representing a similarity between the histogram of the first image and the histogram of the second image;   a unit calculating a spatial correlation image similarity representing a similarity between the spatial correlation image of the first image and the spatial correlation image of the second image; and   a unit determining whether a border between the first image and the second image is a cut change based on the histogram similarity and the spatial correlation image similarity.