JP2010039973A - Method for estimating degree of gaze at each part of image to correct the image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve problems associated with increased definition of image, or increase in various image processing loads and increase in visually stressful images. <P>SOLUTION: The tendency of personal visual sense to gaze at an area with dense irregular color change on an image is used. By estimating the degree of gaze at each part of the image, image processing can be limited to an area with high degree of gaze to reduce the operation load, or an area with low degree of gaze can be highly smoothed to improve the visibility of the image. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

    The present invention relates to image processing for generating an image with high visibility.

There is a method of removing noise by smoothing while maintaining information in the vicinity of an edge of an image (see, for example, Patent Document 1). In the present invention, based on the density of extracted edges, even edges that are not dense are smoothed.
Special Publication 2004-214756

  High-definition images present many details that are not visually important, which increases the load of various image processing, and also causes problems such as visual fatigue and reduced appeal. .

  The present invention makes it possible to estimate an area on the image where a person's attention is directed and to perform image processing according to the degree of gaze. Take advantage of the fact that human vision tends to focus on areas where irregular color changes are concentrated.

  In claim 1, first, a regular color change such as texture is suppressed from the target image. For this purpose, a filter that attenuates main frequency components included in the regular color change is used. Next, an image in which the edge intensity is extracted in each pixel is generated, and this is strongly smoothed, whereby the degree of color change density in each part of the image is quantified. This is a gaze degree estimation diagram.

  According to the second aspect, the gaze degree estimation diagram is used to strongly reduce the information of the region where the attention of the person is not suitable from the target image. For each pixel of the target image, a variable characteristic filter is applied in which the cutoff amount of the high-frequency component increases as the corresponding gaze degree estimation value decreases. As a result, an image with improved visibility can be obtained.

  By using the first aspect and limiting the image processing to a region having a high gaze degree, it is possible to reduce the calculation load.

  The image processed according to claim 2 attenuates a lot of detailed information while maintaining the sharpness of the gaze area. For this reason, vision can be quickly concentrated on the gaze area, and fatigue caused by long-time viewing is reduced.

  The image conversion program is mounted on a computer including an imaging device and a display device. Digital image data obtained from the imaging device is stored in a storage device, and the data is converted by a program process by the CPU and displayed on a display device.

  A configuration for correcting the image by estimating the degree of gaze of each part of the image by program processing by the CPU is shown in FIG.

  In the first aspect, a reduced image having a long side of about 160 pixels is obtained by using low-resolution conversion by a moving average method as means 1a for attenuating regular color change for the target image 0a. As a result, a fine fabric pattern or the like is smoothed to an intermediate color. Next, in the means 1b for obtaining the density distribution of the color change, an edge image obtained by calculating the distance average in the Lab color space with the surrounding pixels in each pixel is generated, and this image is subjected to a Gaussian filter having a variance value σ = 4. Smooth. The distribution of the color variation obtained in this way is referred to as gaze degree estimation FIG. FIG. 3 is a gaze degree estimation diagram obtained by applying this method using the photograph of FIG. 2 as an example of the target image. The high brightness part represents a dense area with irregular color fluctuations.

  According to the second aspect, adaptive smoothing based on the gaze degree estimation value is performed in each part of the original target image 0a. In each pixel, a variable characteristic type Gaussian filter 2a is applied that increases σ as the gaze degree estimation value at the corresponding position is lower. For example, if the target image has 640 pixels on the long side, the σ> 10 is adjusted around the estimated value 0. Thereby, the image 2b with good visibility in which the region with a low gaze degree is strongly smoothed is obtained. FIG. 4 shows an image obtained by applying this method to the photograph of FIG.

  Claim 1 can also be applied to a photographing mechanism of a digital camera. When an area with a high gaze degree is detected in the subject background, the visibility of the subject can be improved by automatically adjusting the depth of field to suppress the surrounding color change.

  According to the second aspect of the present invention, it is possible to enjoy a video that does not get tired if it is incorporated in a video device and a moving image such as a movie or a drama is corrected in real time.

Configuration diagram of processing for correcting the image by estimating the degree of gaze of each part of the image Photo as an example of the target image Gaze degree estimation diagram generated according to the embodiment Photos with improved visibility by example

Claims (2)

  A means for suppressing regular color change with respect to a target image, and means for obtaining a density distribution of color change by edge extraction and smoothing. How to estimate.   A method for generating an image with high visibility, comprising the method according to claim 1 and means for strongly smoothing an area estimated to have a low gaze degree with respect to a target image.