The principle of atitit image definition ambiguity detection and recognition evaluation algorithm

The principle of atitit image definition ambiguity detection and recognition evaluation algorithm

1.1. image Edge is usually achieved by gradient operation of the image 1

1.2. Remark: 1

1.3. 1. Lost focus detection. The Main method to measure the blur is the statistical characteristics of the gradient, the higher the gradient value, the richer the edge information and the clearer the image. 1

1.4. using edge detection , blurred image edges will be less than 2

1.5. Compare by DCT . Comparison of low frequency signals separated by Dct 2

1.6. Reference 2

1.1.image edge is usually achieved by the gradient operation of the image.

1.2. Remark:

1) the quality of the above five images can be distinguished by the naked eye:img42 > Img81 > Img77 > Img29 > img183
     2) the algorithms that are consistent with the subjective perception are:Brenner,Tenengrad,SMD,SMD2, Energy,Entropy,EAV,JPEG,JPEG2
     3) Variance,Vollaththe data obtained by the algorithm is very close, and the image quality cannot be distinguished.
  4) Laplacianin judgingimg29and theimg183, the quality of the two images is very poor

Remark:
1) The naked eye can distinguish the quality of the tablets by:img20 > img228 > Img56 > img152 > Img23 > img215
     2) the algorithms that are consistent with the subjective perception are:Brenner,Tenengrad,Laplacian,SMD2, Energy,JPEG,JPEG2
  3) Vollat,Entropyalgorithm error is more.
  4) SMD,EAVin judgingImg20and theimg228, the quality of both pictures is very good, the naked eye sometimes difficult to distinguish, so this error in an acceptable range.
  5) Variancein judgingimg23and theimg215, the quality of the two images is very poor.

1.3.1. Lost focus detection.  The main method to measure the blur is the statistical characteristics of the gradient, the higher the gradient value, the richer the edge information and the clearer the image.

The main performance of the loss of focus is blurred picture, the main method to measure the blurred image is the statistical characteristics of the gradient, usually the higher the gradient value, the richer the edge of the picture, the more clear the image. Note that the gradient information and each of the characteristics of the video itself has a relationship, if the picture itself is very small, even if the focus, gradient statistics will be very few, the monitoring equipment lost focus detection needs manual participation in the calibration process, by the person to tell the computer a device under normal conditions of the texture information is how.

1.4. with edge detection , blurred image edges are less

For example, the lower the number of stars, the lower the compression rate, the smaller the picture size, the worse the picture quality. As you can see, the relative edges of the pictures with fewer stars are more blurred. Of course, at a certain compression rate, the naked eye is unable to directly detect the reduction of picture quality (such as Samsung and four stars).

1.5. comparison by DCT . Comparison of low frequency signals isolated from Dct

blur picture less detail, so the DCT is lower.

1.6. References

definition Evaluation method for non-reference images - Lingfeng's Column - Blog channel -CSDN.NET.html

video sharpness, color offsets, and brightness anomaly detection -lengwuqin 's Column - Blog channel -CSDN.NET.html

camera lost focus detection -lien0906 's Column - Blog channel -CSDN.NET.html

detection algorithm for image signal missing or sharpness -Qingkongyeyue 's blog - Blog channel -CSDN.NET.html

author::  Nickname :Old Wow's claws( Full Name::AttilaxAkbar Al Rapanui Attilaksachanui) 

Kanji Name: Etila ( Ayron) , email:[email protected]

reprint Please indicate source: http://www.cnblogs.com/attilax/

Atiend

Atitit image sharpness detection and recognition evaluation algorithm principle of ambiguity