JP2000069297A - Method for objectively evaluating image quality of irreversibly compressed image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To create an image quality evaluation system, in which an arbitrary part such as a ratio of weighting to lock down on a high frequency component is excluded and image quality is evaluated impartially between different image compression systems. SOLUTION: First precision of a color estimated from an error between an original image after smoothing and a compressed image is obtained for every parameter of various smoothing filters. Then a ratio of estimated and stored frequency components is obtained from a ratio of a variance of pixels of the original image to that of the image after smoothing. A maximum value of a product of two indices is used for an evaluation result of the image quality.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[Technical field of use] The process of optimizing the compression format and compression parameters for each image requires a great deal of labor to perform manually, but it is difficult to use video games, multimedia C
At present, designers such as D-ROMs and Webs take time and work on their processes and rely on experience. Automating this allows for labor savings in all fields dealing with digital images. It can be used as an elemental technology in situations such as image search and image compression optimization.

[0002]

2. Description of the Related Art Conventionally, it has been fundamental to obtain the sum of squares of an error. Improvements have been made such as using a uniform color space to make the image more faithful to the eyes, weighting high-frequency components, using correlation instead of the sum of squares of errors, and supporting pseudo-gradation display. But how much do you cut the highs? The difference caused by the advantages and disadvantages of the evaluation method is larger, and it has been impossible to capture the improvement in the image quality due to the slight difference in the compression method and the compression parameter.

[0003]

The sensitivity of human vision decreases as the frequency becomes higher. However, in many cases, the distance from which the image is viewed is undefined. You lose grounds. General-purpose and objective image quality evaluation must be possible without increasing the arbitrariness of the evaluation method.

[0004]

[Means for Solving the Problem] A unique evaluation independent of parameters in an evaluation method is achieved by a mechanism for optimizing a parameter which may enter arbitrariness so as to maximize the amount of information held in a compressed image. The result was obtained.

[0005]

The more the high frequency is cut by the smoothing filter, the more the difference between the image before and after the compression becomes smaller and the color depth becomes larger, but on the other hand, the frequency component of the image is lost more and more, However, the entire screen is filled with one color constant. In this image quality evaluation method, a product of a color depth and a ratio of a frequency component is regarded as image quality. In the evaluation of a compression method, a compression method that maximizes image quality per file size is selected as a good compression method. As a result, for example,
It has become possible to objectively evaluate a compression method in which high-frequency components are discarded, such as JPEG, and a method, in which high-frequency components are preserved, such as color reduction, in a common playing field that is not disadvantageous to either.

[0006]

DESCRIPTION OF THE PREFERRED EMBODIMENTS (1) Main Loop A routine for estimating the ratio of the amount of information to be stored while changing a variance as a parameter using a two-dimensional normal distribution as a smoothing filter is repeatedly called. (2) Estimation of Color Depth The original image and the compressed image are respectively passed through a smoothing filter, and then the sum of squares of the error is obtained. By comparing this with the original image that has passed through the smoothing filter, and the sum of squares of the error when comparing the image with the lower bits truncated,
Estimate color depth. (3) Estimation of Stored Frequency Components The ratio of the variance of all pixels of the original image to the variance of all pixels of the original image passed through the smoothing filter is used as the ratio of the stored frequency components. (4) The product of the color depth and the variance ratio is obtained, and the maximum value is stored.

[0007]

As described above, when a new compression method is developed or improved, the development efficiency is increased by the presence of an objective and unique index. Image reduction can also be called irreversible compression of images, but when tagging images for search, the amount of image information per file size is used to determine the size of the image before extracting the tags. Can be uniquely determined by using an index such as maximizing. Further, by dividing an image into blocks, a part having a small number of colors for each block uses GIF, and a gradation part can be used as a discriminating part of a general-purpose image compression method combining advantages such as using JPEG.

[Brief description of the drawings]

FIG. 1 is a schematic flowchart of an embodiment.

Claims (1)

[Claims] In an objective image quality evaluation, an original image and an irreversible compressed image are passed through a smoothing filter having the same parameters, and then a sum of squares of errors is obtained to estimate a color depth (accuracy).
On the other hand, the ratio of the stored frequency component is determined by comparing the original image and the image obtained by smoothing the original image. This process is repeated for various smoothing filter parameters,
A method using the maximum value of the product of the color depth and the storage ratio of the frequency component as the image quality.